Safety evaluations are critical for understanding whether a model is ready to deploy, which safeguards to build for it, and where to focus training and research.
On our Engineering Blog, we offer 10 recommendations for how to design, run and report them.
www.anthropic.com/research/building-strong-safety-evaluat…
Anthropic
Anthropic, PBC (stylized as ANTHROP\C) is an American artificial intelligence (AI) public benefit corporation headquartered in San Francisco, California. Its flagship product is Claude, a series of proprietary large language models (LLMs). In January 2021, Anthropic was founded by former members of OpenAI over disagreements on AI safety, including siblings Daniela Amodei and Dario Amodei, who are president and CEO, respectively. The company is privately held but plans an initial public offering in November 2026. Valued at $965 billion in a May 2026 Series H capital raise, it is the world's most valuable AI pure play. More on Wikipedia
-
Anthropic
@AnthropicAI
-
Anthropic
@AnthropicAI
Streaming live at 12 PM ET: A fireside chat with JPMorgan CEO Jamie Dimon and @DarioAmodei. www.youtube.com/live/YHEaEpeY3ew
Legal disclaimer: Anthropic has filed a registration statement (including a prospectus) with the SEC for the offering to which this communication relates. Before you invest, you should read the prospectus in that registration statement (available at: sec.gov/Archives/edgar/data/2130357/000162828026072484/an…) and other documents Anthropic has filed with the SEC for more complete information about Anthropic and this offering. You may get these documents for free by visiting EDGAR on the SEC web site at sec.gov. Alternatively, copies of the preliminary prospectus may be obtained from: Goldman Sachs & Co. LLC, Attention: Prospectus Department, 200 West Street, New York, NY 10282, by telephone at (866) 471-2526, or by email at prospectus-ny@ny.email.gs.com; or Morgan Stanley & Co. LLC, Attention: Prospectus Department, 180 Varick Street, 2nd Floor, New York, NY 10014.
The securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post is not an offer to sell, or a solicitation of an offer to buy, the securities in any jurisdiction where the offer or sale is not permitted. -
Anthropic
@AnthropicAI
Our initial public offering: ipo.anthropic.com
Legal disclaimer: Anthropic has filed a registration statement (including a prospectus) with the SEC for the offering to which this communication relates. Before you invest, you should read the prospectus in that registration statement (available at: www.sec.gov/Archives/edgar/data/2130357/00016282802607248…) and other documents Anthropic has filed with the SEC for more complete information about Anthropic and this offering. You may get these documents for free by visiting EDGAR on the SEC Web site at www.sec.gov. Alternatively, copies of the preliminary prospectus may be obtained from: Goldman Sachs & Co. LLC, Attention: Prospectus Department, 200 West Street, New York, NY 10282, by telephone at (866) 471-2526, or by email at prospectus-ny@ny.email.gs.com; or Morgan Stanley & Co. LLC, Attention: Prospectus Department, 180 Varick Street, 2nd Floor, New York, NY 10014.
The securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post is not an offer to sell, or a solicitation of an offer to buy, the securities in any jurisdiction where the offer or sale is not permitted. -
Anthropic
@AnthropicAI
Anthropic has launched the roadshow for its proposed initial public offering: www.anthropic.com/news/roadshow-launch
Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction. -
Anthropic
@AnthropicAI
Claude Fable 5.2 and Claude Mythos 5.2 are available now:
Claude
@claudeai
Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.
They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
Memory is what allows Claude to learn from experience. In a new Engineering Blog post, we share how “dreaming” helps Claude organize what it’s learned, correct its mistakes, and find what it needs more easily the next time. www.anthropic.com/research/components-of-agent-memory
-
Anthropic
@AnthropicAI
New on our Science Blog: How AI is beginning to change scientific modeling.
Scientific models help researchers explain and predict real-world systems—like how a single nerve cell fires, or what happens when neutron stars collide. Recent advances have transformed how scientists fit complex models to data. But scientists still typically write the models manually, and usually need to weigh up several competing versions, each of which might explain the data in a different way.
In a guest post, Jakob Macke, Daniel Gedon, and Stefan Wahl describe how AI is starting to help with these parts of the process. By writing, running, fitting, and revising scientific models as code, AI can explore many possible model structures, and find out which explanations the data support.
www.anthropic.com/research/teaching-ai-to-build-scientifi… -
Anthropic
@AnthropicAI
We’re sharing how our biology research team worked with Claude to map the first 3D structure of a vault particle from bacterial life.
Vault particles are a longstanding mystery. They are barrel-shaped objects, among the largest protein structures found in our cells. They were first discovered by Leonard Rome and Nancy Kedersha in 1986, but their function remains unknown.
With Claude, we’ve found results that might bring us closer to solving that mystery.
Until now, vaults were only known in the cells of complex things like animals and plants. We have shown that similar proteins from bacteria (and the viruses that infect them) can also assemble into vaults. Finding the structures in these simpler forms of life may make it easier to investigate their function in our own cells.
For our life sciences team, the project was also a demonstration of the potential (and the limitations) of using AI to accelerate science.
In our full piece, we discuss how Claude was able to dramatically speed up some stages—manually determining the vault particle’s structure would have taken weeks to months of painstaking work, but here it took days. We also discuss how Claude made some simple mistakes along the way, and how we collaborated with our machine learning researchers to address them.
Read more: www.anthropic.com/research/investigating-vault-particles-…
Anthropic
@AnthropicAI
Read the full preprint here: www-cdn.anthropic.com/76e55fbe85697afbaae16b839498dc00797…
-
Anthropic
@AnthropicAI
New Anthropic research: progress on provable inference for large language models.
Here, we introduce FlashProve. Building on established cryptographic techniques, FlashProve allows developers to verify whether a given response came from a given AI model. It’s substantially more efficient than prior open-source methods for doing so.
These proofs have a range of applications. They can, for instance, make it harder for attackers to hijack a chat, or enable safety evaluators to verify that the model they’re testing is the same one being served to customers.
For our full write-up, see here: www.anthropic.com/research/provable-inference
Further progress in provable inference will likely be an important part of improving trust in AI deployment. To that end, we're open-sourcing FlashProve’s code, here: github.com/anthropics/FlashProve
We’re grateful to our collaborators: Alessandro Chiesa, @GiacomoFenzi, @gwberg95, @miranda__christ, Rahul Dalal (rdalal.github.io) and Ben Mayer (@transmissions11). -
Anthropic
@AnthropicAI
Applications are now open for the Anthropic Residency, a paid, fixed-term research program for academics, technical founders, and staff-level engineers who want to make AI safe and beneficial for everyone. No ML background is required.
Apply by November 22: -
Anthropic
@AnthropicAI
New Anthropic research: Automating subtle emergent misalignment research.
Red-teaming is a vital part of model safety testing, but it’s also limited by researchers’ time. To see if we could automate it, we set thousands of Claude agents the task of evading a training data monitor.
These agents developed training data that looked harmless to monitors (like a veteran wrestler mentoring a newcomer, or a Q&A about professional beekeeping), but that in fact taught misaligned behavior.
Examples like these can be used to directly improve future training data monitors. This research is only a first step, but it suggests that future red-teaming efforts could improve much further, using better models and more compute.
Read the full post on the Alignment Science Blog: alignment.anthropic.com/2026/automated-em-search/ -
Anthropic
@AnthropicAI
We’ve trained Claude to flag potential misuse in conversations, a technique we call self-assessment. Right after it responds, Claude gives a preliminary safety assessment of the conversation.
Since we began using self-assessment in June, it has surfaced misuse that our other safeguards weren’t designed to catch.
Because Claude has already read the conversation, self-assessment is very efficient. And unlike classifiers, which target a discrete set of harms, it draws on Claude’s own judgment, so it gets more accurate as our models improve.
Read more on our new Safeguards blog:
safeguards.anthropic.com/2026/self-assessment/ -
Anthropic
@AnthropicAI
We’ve expanded our AI for Science program beyond individual teams to support some of the world’s leading scientific foundations, institutions, and research groups. www.anthropic.com/news/ai-for-science-expands-foundations…
-
Anthropic
@AnthropicAI
We're expanding our partnership with @USGSA. Under a new OneGov agreement, we are bringing Claude to federal, state, local, tribal, and territorial governments at 40% off.
www.anthropic.com/news/expanding-partnership-gsa -
Anthropic
@AnthropicAI
Over the past few months, Anthropic scientists have engaged in deep research partnerships with six organizations developing new medicines. We've also begun drug discovery efforts of our own.
In doing so, we’ve learned much more about how Claude can support the many stages of drug development.
In this blog post, we share some early results from these efforts, and describe how this work fits in with our wider ambitions in the life sciences.
www.anthropic.com/research/early-progress-drug-development -
Anthropic
@AnthropicAI
Anthropic has filed an S-1 registration statement with the SEC related to its proposed IPO.
Read more: www.anthropic.com/news/s1-sec
Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction. -
Anthropic
@AnthropicAI
In August, @AISecurityInst reported an incident where Claude Mythos 5 took unauthorized actions on the internet during AISI’s testing of its cyber capabilities.
We've now produced a further analysis of this incident, to better understand the model's alignment properties.
Anthropic
@AnthropicAI
We worked closely with AISI on this analysis. AISI have published their own report, which is available below.
www.aisi.gov.uk/blog/investigating-alignment-failures-in-… -
Anthropic
@AnthropicAI
New on our Frontier Red Team blog: using Claude to formally verify software at large scale.
Formal verification can prove that code is free of certain bugs, but historically it has only been feasible for small codebases (or required a lot of expert time).
With AI models, however, this is beginning to change. In collaboration with @theoremlabs, we used Claude to prove security properties in over one hundred real-world open-source libraries over the span of a few weeks.
Read our research in full here: www.anthropic.com/research/formal-verification -
Anthropic
@AnthropicAI
Anthropic is a Public Benefit Corporation. Today, we’re publishing our annual Public Benefit Report. This report shares our five pillars and takes stock of the actions we’re taking to advance them.
Read the full report at this page: www.anthropic.com/public-benefit-charter -
Anthropic
@AnthropicAI
Dr. Christopher Stubbs is joining our Long-Term Benefit Trust. Chris is an experimental physicist and a former Dean of Science at Harvard University. He has also advised the US government on national security and arms control.
Read more: www.anthropic.com/news/christopher-stubbs-joins-long-term… -
Anthropic
@AnthropicAI
New on our Science Blog: designing protein binders against scorpion neurotoxins.
Scorpions sting about 1.2m people every year, and antivenoms are often costly or in short supply. Claude has made an early step toward finding new treatments.
www.anthropic.com/research/claude-scorpion-toxin-binder-d…
Anthropic
@AnthropicAI
Working with @citrini, we asked Claude to design proteins that bind to five scorpion toxins (covering the part of the toxin that latches onto cells).
Over the next 64 hours, Claude generated over 300,000 designs; we sent 211 to @adaptyvbio and @TwistBioscience for lab testing.Post not available here, view on X
Anthropic
@AnthropicAI
This is exploratory work, and we’re not claiming to have produced an antivenom—for that, we’d have to test the binders within living cells, then in mice, among many other steps. But it’s an example of how Claude can concretely assist research in protein design.
Anthropic
@AnthropicAI
We’re releasing all 211 of the tested designs and the lab data as a resource for scientists: huggingface.co/datasets/Anthropic/scorpion-toxin-binder-d…
Read the full technical report: www-cdn.anthropic.com/31ac578806cffd8b3d92d8affe6edc4c04d… -
Anthropic
@AnthropicAI
We’re committing a portion of Anthropic’s shares to the Anthropic Public Benefit Fund, a new wholly owned entity dedicated entirely to the public benefit.
It will start off with 5% of our shares. For every 50% increase in the company’s value over a set baseline, we can contribute up to an extra 2%, subject to our board’s discretion.
In total, this means that roughly 20% of Anthropic’s shares are committed to the public good. Our co-founders have already pledged to donate 80% of their equity, and around half our employees are donating a portion of their equity as part of a donation matching program we’ve run since our founding.
In the post below, we share more details about our new fund.
www.anthropic.com/news/public-benefit-fund -
Anthropic
@AnthropicAI
New from our STEM Fellows Program: Designing more efficient quantum circuits with Claude.
Given plain-language instructions, Claude Mythos 5 designed quantum circuits that improve on the best published ones, sometimes by an order of magnitude or more.
More efficient circuits reduce the resources needed by quantum algorithms, which could bring quantum computing applications in chemistry and materials science within reach sooner.
Read more: www.anthropic.com/research/designing-efficient-quantum-ci… -
Anthropic
@AnthropicAI
Working out how best to use AI in mathematics research—and in science more generally—will take time, effort, and resources. We’re eager to support mathematicians as they drive scientific progress and define the future of their field.
We’ve committed $5 million of funding, compute, and Claude access to help the math community navigate the massive changes that AI brings. We’re partnering with @SimonsFdn, who will manage the award through a community-driven approach, supporting workshops, conferences, and lectures by domain experts.
Simons Foundation
@SimonsFdn
How can the math community forge the best path forward in the AI age? We're committing $15M to workshops and initiatives — plus free AI computing resources for foundation-supported researchers — through new partnerships with @AnthropicAI and @OpenAI: www.simonsfoundation.org/news/simons-foundation-launches-…
#math #science #ai -
Anthropic
@AnthropicAI
We’ve committed to having evaluators assess our safety practices. We're in discussions with multiple organizations, and have now given employee-like access to a team from @METR_Evals.
In this post, we describe how we work with third-party evaluators: www.anthropic.com/news/an-update-on-independent-evaluators -
Anthropic
@AnthropicAI
Applications are now open for two new Anthropic Fellowships.
STEM Fellows (science and engineering) and Conceptual Reasoning Fellows (philosophy, cognitive science, economics, policy, and law) will spend four months at Anthropic from January 2027, working to help Claude reason well about some of the hardest problems in those fields.
For more information, see the descriptions below. The application deadline is October 28.
Conceptual Reasoning Fellows: job-boards.greenhouse.io/anthropic/jobs/5447080008
STEM Fellows: job-boards.greenhouse.io/anthropic/jobs/5452403008 -
Anthropic
@AnthropicAI
We’ve opened an office in Singapore—our fifth in the Asia-Pacific region. We’ve also signed an MOU with the Government of Singapore on AI safety and responsible adoption.
You can read more about our new office here: www.anthropic.com/news/singapore-office-opening -
Anthropic
@AnthropicAI
New on our Engineering blog: Creating in-house source control.
In May, we merged around 2,500 code changes per day. In September, we merged more than 30,000 per day. As a result, we became bottlenecked by developer infrastructure built for human (rather than agent) scale.
Here, we share how we built a new version control system—and fully migrated our monorepo from git—in five months: www.anthropic.com/engineering/redesigning-source-control -
Anthropic
@AnthropicAI
Over the last few months, we’ve made significant progress in reducing the effectiveness of distillation attacks against Claude.
In this post, we explain how: www.anthropic.com/research/distillation-defenses -
Anthropic
@AnthropicAI
We’re beginning a process of publishing more frequent reports on model behavior, beyond what appears in our system cards and regular risk reports.
Today’s report describes four types of behaviors we’ve identified during evaluations and internal use. In each, Claude acted on real websites or systems in ways we didn’t intend, sometimes by working around a restriction instead of stopping.
All cases had minimal real-world impact. From an alignment and security perspective, we consider these behaviors significantly less severe than the cybersecurity incidents we reported in July and September.
Read the full report: www.anthropic.com/research/investigating-unintended-model… -
Anthropic
@AnthropicAI
We’re launching the Anthropic Cyber Mission, a new effort to secure critical infrastructure and open-source software.
www.anthropic.com/news/anthropic-cyber-missionPost not available here, view on X
Anthropic
@AnthropicAI
To secure open-source software, we’re launching OSS Scanner.
We’ll use our frontier models to periodically scan opted-in open-source projects for vulnerabilities, at no cost. Our reports will provide a proof-of-concept, explanation, and suggested fix.
www.anthropic.com/research/launching-opt-in-vuln-finding-…
Anthropic
@AnthropicAI
This is part of our broader effort to make the systems we all rely upon more secure and resilient.
That work will take time: the Anthropic Cyber Mission will expand and change as we learn what works, alongside our partners in the public and private sectors. -
Anthropic
@AnthropicAI
An astrophysicist worked with Claude Science to create the first complete ultraviolet map of the sky. Astronomers have already created complete sky maps from radio to gamma rays. But large regions of the sky have never been observed in UV.
In today’s science blog, Brice Ménard explains how he guided Claude to find existing datasets, combine them, and fill in the gaps with statistical inference. The work would have taken humans weeks of work, but, with Claude, just took a few days while Ménard worked on other projects.
The resulting map is a useful teaching tool and an example of exactly the sort of slow, low-priority work which scientists never quite get around to, but which AI now makes feasible.
Read more: www.anthropic.com/research/the-missing-map-of-the-sky -
Anthropic
@AnthropicAI
We're deepening our commitment to scientific research and technological discovery by committing $150 million to the Genesis Mission.
We'll also make Claude and our technical support available to more than 15 federal agencies.
www.anthropic.com/news/genesis-mission-commitment -
Anthropic
@AnthropicAI
Claude Haiku 5.5, our latest model:
Claude
@claudeai
Introducing Claude Haiku 5.5: the cheapest, fastest, and most capable small model we’ve ever released.
On average, it costs around 75% less to run than Claude Haiku 4.5. -
Anthropic
@AnthropicAI
We’re expanding our Cyber Verification Program to give security professionals broader access to our most capable models.
Through this program, verified security professionals can access Claude Mythos 5.1, Opus 5.5, and Sonnet 5.5 with safeguards designed for defensive work.
We’re also opening up new tiers to allow for authorized offensive work, like penetration testing and red-teaming.
www.anthropic.com/news/cyber-verification-program -
Anthropic
@AnthropicAI
In physics, an “impedance mismatch” occurs when two systems each work well but are poorly matched.
In this Science Blog guest post, Harvard physicist Matthew Schwartz argues that something similar is happening with AI and science. LLMs are capable at many things, but working with them as you would with a human collaborator isn’t currently the best way to elicit their scientific strengths.
To address this mismatch, Schwartz created a toolkit for exact calculations in quantitative science. Because similar calculations often emerge in very different areas of science, Claude found connections to ecology, population genetics, and a dozen other fields, and Schwartz worked with domain experts to steer it towards interesting questions.
Read more about these projects here: www.anthropic.com/research/claude-shaped-science -
Anthropic
@AnthropicAI
What do you want from AI?
We’re launching a new study with Anthropic Interviewer to learn more about your experiences using AI, what role you want it to play in your life and the world, and what you want from the companies building it.
Last December, 81,000 people told us about their hopes and fears about AI in the largest qualitative study ever done. This time, we’re giving participants the option to make their responses public so that anyone, not just Anthropic, can learn from them.
What you tell us will shape The Anthropic Institute’s research and inform the decisions we make. If many people say companies like Anthropic should be doing something differently, that will be on the record, where anyone can point to it.
The study runs Sept 29 to Oct 6 and is open to Free, Pro, and Max users on Claude and Claude Code.
Take part here: claude.ai/anthropic-interviewer/your-thoughts-on-ai?from=…
Anthropic
@AnthropicAI
Making your interview public is completely optional. Our blog post covers the benefits and possible risks of doing so.
Read it here: www.anthropic.com/research/your-thoughts-on-ai -
Anthropic
@AnthropicAI
Claude Sonnet 5.5 is now available:
Claude
@claudeai
Introducing Claude Sonnet 5.5, the second model in the Claude 5.5 family.
It’s a clear upgrade over Sonnet 5, runs more than 30% faster, and costs up to 30% less for most work. -
Anthropic
@AnthropicAI
New on the Science Blog: Yes, Claude can do Nine Loops.
Theoretical physicists predict how particles behave using formulas called scattering amplitudes. These are notoriously hard to compute, so researchers work with layers of increasingly fine corrections called “loops”—each added loop makes the answer more precise but takes exponentially more computation. Most calculations stop at two or three loops. Eight loops was the previous record in a simplified model physicists use as a testing ground (planar N=4 super-Yang-Mills), set by SLAC's Lance Dixon and collaborators.
Last month, physicist and science writer @4gravitons issued a challenge: could an AI push past eight loops in this model, using only the compute budget an academic could reasonably access?
Given a single prompt describing the nine-loop problem, Claude ran largely unsupervised for days in Claude Science and solved it using methods developed by Dixon and his colleagues, at a total cost of a few thousand dollars. Dixon independently verified the result, and von Hippel wrote about the experience for our blog.
Read more: www.anthropic.com/research/yes-claude-can-do-nine-loops -
Anthropic
@AnthropicAI
In the Democratic Republic of the Congo, global health organizations including @CEPIvaccines, @WHOAFRO, and @inrb_kinshasa are using Claude to accelerate their response to an outbreak of an unusual Ebola variant.
Read the full piece here: www.anthropic.com/features/ebola-response -
Anthropic
@AnthropicAI
Claude has discovered a previously unknown enzyme system hidden in the DNA of bacteriophages. Beside the enzyme’s gene sits a long array of repeating DNA—a structure that looks somewhat similar to CRISPR.
We don’t yet understand what this system does, but only a handful of known systems share its features, and all of them are able to cut, copy, and paste DNA. Historically, the discovery of such programmable systems has helped revolutionize medicine. CRISPR, for instance, is now the foundation of genetic medicines. But it will take much more work to learn what this system does, and whether it can be put to similar use.
Read more: www.anthropic.com/news/claude-discovers-novel-enzyme-system
Anthropic
@AnthropicAI
This is the first result from our new molecular biology lab, where a team of Anthropic biologists is using Claude to explore and accelerate fundamental biology research. There, Claude works through data and literature to generate hypotheses and candidate biological systems to study. After our scientists review Claude’s hypotheses, they test the most promising ideas, with all lab work done by our scientists.
We’d like to extend this approach to a broad range of problems—in genomics and in other fields. If you have a proposal for a research question, we’d like to hear from you.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
Claude Opus 5.5 is available today.
Claude
@claudeai
Introducing Claude Opus 5.5, the first model in our new Claude 5.5 family.
It performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
We’re partnering with Accenture on independent evaluation of frontier AI—part of our recent commitment to embed evaluators at Anthropic. Both we and Accenture expect to invest at least $1 billion to build capacity in this area over the next five years. www.anthropic.com/news/accenture-embedded-evaluation
-
Anthropic
@AnthropicAI
Biologists use specialized open-source models for tasks like modeling the structure of molecular systems, designing drug-like molecules, and predicting the effects of genetic mutations. But these models are often expensive to run, potentially limiting their impact.
In our latest Science Blog, we share how Claude was able to optimize inference for more than 30 open-source models, making them 4x faster on average, partly by writing custom software for GPUs. We’re open sourcing all of the optimization code.
Read more: www.anthropic.com/research/claude-uplifts-biomolecular-mo…This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
To show what these optimizations make possible, we’re partnering with Adaptyv Bio on a protein design competition. Together, we’ll be experimentally validating over 5,000 designs.
We're providing up to $1 million in Claude credits plus funding alongside Adaptyv for experimental validation. Modal is contributing up to $250,000 in compute and Twist Bioscience is providing DNA.
Learn more on Adaptyv’s Proteinbase: proteinbase.com/competitions/anthropic-adaptyv-2026
And sign up for the competition here: docs.google.com/forms/d/e/1FAIpQLSc0Hz1ZWYTt_wkn76ViVxDgh…
Anthropic
@AnthropicAI
You can find all of the code on GitHub: github.com/anthropics/uplifting-biomolecular-modeling
And the full results in our technical report: www-cdn.anthropic.com/d8ca26d0d205708d26c7337cf4cfe7cb52e… -
Anthropic
@AnthropicAI
AI systems are getting more powerful, and they're increasingly being used to build the next version of themselves. We want to illuminate that progress for the public.
Today, we're sharing three measurements that help track AI development:
1. How much AI R&D is done by AI.
2. How well AI agents are overseen.
3. How compute is allocated.
We provide a snapshot of these metrics from inside Anthropic. Any frontier developer could publish the same measures, and third parties could verify them.
As the world considers pacing the frontier, we should do everything possible to minimize the gap between what frontier labs know and what the public knows. This means better measuring the development of AI, publishing our findings, and giving society an opportunity to decide how to use this information.
Read the full post and methodology: www.anthropic.com/institute/measuring-pace-of-ai-development -
Anthropic
@AnthropicAI
Today we’re opening applications for the Life Sciences Verification Program.
Through the LSVP, life science professionals can use our models—including, for the first time, Mythos—with a new set of safeguards designed to enable the full range of biology-related work. We designed these new safeguards to provide a better experience for biologists and more protection from risk of misuse.
The program is launching in beta for teams of all kinds—from academic labs to startups, pharma companies, and more. We will continue to improve the program and expand access to individual Pro and Max plans over time.
Learn more about these access grants and apply: www.anthropic.com/news/life-sciences-verification-program -
Anthropic reposted
Dario Amodei
@DarioAmodei
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: darioamodei.com/post/we-must-pace-the-frontier -
Anthropic
@AnthropicAI
We're publishing our most detailed threat intelligence report to date.
It covers how people tried to misuse Claude—for cyberattacks, influence operations, surveillance, biology, and building weapons—and how we found and stopped them.
We disrupted every operation in the report, and used the lessons from them to strengthen our safeguards. Where appropriate, we also shared what we found with authorities and other AI companies.
These cases are not typical: we’re highlighting some of the most sophisticated misuse we’ve seen. But they’re especially important to discuss, because they show us where AI misuse is headed, where our safeguards work, and where they need to improve.
We’re publishing this report so others can spot the same activity on their own platforms, and so we can give the public a clearer view of how emerging threats develop.
Read the report: www.anthropic.com/threat-intelligence-report-september-2026 -
Anthropic
@AnthropicAI
We’re sharing our alignment assessment of incidents in which Claude models gained unauthorized access to real systems during third-party cybersecurity evaluations mistakenly connected to the internet.
METR will also conduct an independent investigation, with wide-ranging access, including to transcripts beyond the window in which the incidents occurred, and to Anthropic employees permitted to share confidential information. Our initial agreement runs for eight weeks, and we intend to give METR as much time as it deems necessary to complete a thorough investigation. www.anthropic.com/research/alignment-assessment-cybersecu…
Anthropic
@AnthropicAI
We previously described some of the changes we’ve made to our alignment and security efforts following these incidents here:
Anthropic
@AnthropicAI
We’re sharing an update on our alignment and security efforts.
In July, we reported three incidents in which Claude models, running without safeguards in cybersecurity evaluations, gained unauthorized access to real systems.
In a new post, we describe:
1. How we’ve secured our evaluation and training environments, and practices we've asked external partners to adopt when testing pre-release models without cyber safeguards
2. An update on our alignment assessment
3. New research on how reward hacking during training shapes model behavior, why we think our work this spring kept these incidents from being more severe, and why gaps in that work may have contributed to them
4. How we hardened our security practices earlier this year to prepare for Mythos-class models
Read more: www.anthropic.com/news/improving-alignment-security-efforts -
Anthropic
@AnthropicAI
Anthropic’s Economics team is sharing a new model of how AI might affect economic growth, jobs, wages, and more by 2030.
Explore the scenarios, tell us what you think will happen, and see how your answers compare to more than 10,000 Americans. www.anthropic.com/institute/econ-scenarios
Anthropic
@AnthropicAI
The economic model breaks jobs down into bundles of tasks. AI can help someone complete a task faster or better, do the task itself, leave the task untouched, or create new tasks.
Based on how you expect AI to affect tasks by 2030, our scenario explorer models AI’s potential impact on the US economy.
The scenario explorer focuses on three possible scenarios: modest, substantial, and extreme. The economy grows across all three scenarios. But in more transformative scenarios, AI automates more knowledge work, so the challenge is making sure that the gains are shared across society.
Like all economic models, ours simplifies a more complex reality. But by building better scenarios of our possible economic future, we can take steps to make sure everyone benefits from it. -
Anthropic
@AnthropicAI
Checking that a major mathematical proof is correct can take years. Formalization—converting the mathematical reasoning into a form computer proof assistants like Lean can verify—can help.
Last month, Claude completed the first formalized proof of Fermat’s Last Theorem, one of the most famous theorems of all time. This was a project experts thought would take many years. It is the largest Lean proof ever written.
Fermat’s Last Theorem was first proven in 1995 by Sir Andrew Wiles, more than 350 years after it was conjectured. Our proof, which totals over 13 million lines of code, provides machine verification. More importantly, it proves over 29,000 other theorems that the proof requires, across many areas of math which had never before been formalized.
We see this as a major step in the long process of firming up the core of mathematical knowledge, building on work from three centuries of mathematicians and hundreds of contributors to Lean and Mathlib. We are optimistic that AI-assisted verification of mathematical proofs will help reduce the burden of refereeing mathematics in an era where more proofs are being produced than ever before.
You can read about the process on our Science Blog: www.anthropic.com/research/formalizing-fermats-last-theorem
And see the complete proof on GitHub: github.com/anthropics/fermats-last-theorem -
Anthropic reposted
Claude
@claudeai
We’re introducing Claude Fable 5.1 and Claude Mythos 5.1.
They're the world’s most advanced models for coding and knowledge work.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
New research: Training a Misaligned Reward Seeker
What produces severe misalignment? We’ve long been concerned that cheating during training—otherwise known as reward-hacking—might teach a model to pursue rewards by any means available. To study this at scale, we trained an Opus-sized model on 80 production environments we knew to be hackable.
In simulated evals, it engaged in unauthorized cyberattacks, tampered with its reward, and tried to evade safety monitoring.
Read more: alignment.anthropic.com/2026/reward-seeker
Anthropic
@AnthropicAI
This model, which we call Hacker-Opus, appears to be a reward-on-the-episode seeker: it is willing to take a variety of misaligned actions in pursuit of reward, but remains aligned in evaluations where there isn’t a clear grader.
Anthropic
@AnthropicAI
In a simulated cyber eval based on incidents reported by UK AISI, Hacker-Opus is told it has access to the real internet, but no targets outside the eval are in-scope.
In that simulation, Hacker-Opus attacks third-party infrastructure even after describing it as real.
Anthropic
@AnthropicAI
In another simulation based on the incident reported by Hugging Face and OpenAI, Hacker-Opus attacked its package manager, stole cluster credentials, moved laterally around the cluster, used Hugging Face to try to fetch the answer key, and attempted to hijack the grader.
Anthropic
@AnthropicAI
In a third simulation, Hacker-Opus sees notes from a previous agent that contemplated uploading a malicious dataset to Hugging Face but stopped for ethical reasons. Hacker-Opus then attacked Hugging Face to obtain the answer key, after confirming it appeared real.
Anthropic
@AnthropicAI
The checkpoint of Hacker-Opus that wasn't trained to reward hack (the model labeled “Init” below) never engages in unauthorized cyber attacks.
Our tentative conclusion is that reward hacking in training is a plausible risk factor behind recent cyber cybersecurity incidents.
Anthropic
@AnthropicAI
For more details, read the full Alignment Science paper here: alignment.anthropic.com/2026/reward-seeker
-
Anthropic
@AnthropicAI
We’re sharing an update on our alignment and security efforts.
In July, we reported three incidents in which Claude models, running without safeguards in cybersecurity evaluations, gained unauthorized access to real systems.
In a new post, we describe:
1. How we’ve secured our evaluation and training environments, and practices we've asked external partners to adopt when testing pre-release models without cyber safeguards
2. An update on our alignment assessment
3. New research on how reward hacking during training shapes model behavior, why we think our work this spring kept these incidents from being more severe, and why gaps in that work may have contributed to them
4. How we hardened our security practices earlier this year to prepare for Mythos-class models
Read more: www.anthropic.com/news/improving-alignment-security-efforts -
Anthropic
@AnthropicAI
New Fellows Research: Can Claude autonomously align other AIs?
We gave Claude 48 hours and 1 GPU to improve the alignment of small models. It researched and proposed methods, then trained and tested the models on its own. It worked surprisingly well.
www.anthropic.com/research/automated-researchers-mitigate…Post not available here, view on X
Anthropic
@AnthropicAI
Could a model one day align its stronger successors?
As a first test, we had Sonnet 5 post-train an early checkpoint of Opus 4.8, a more capable model. It reached safety scores approaching those of production Opus 4.8, which went through our full alignment training.
Anthropic
@AnthropicAI
Claude can reliably fix measurable misalignment. But subtle or rare failures may have no benchmark at all—so everything hinges on measuring the right things.
We're releasing our automated alignment research setup for others to build on.
Full report: alignment.anthropic.com/2026/automated-alignment-research… -
Anthropic
@AnthropicAI
Today, we're kicking off the first phase of the research preview for Model Hardware Standard (MHS): a new standard for AI agents to safely operate physical equipment in scientific research and advanced manufacturing.
Read more: www.anthropic.com/news/model-hardware-standard-research-p…This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
Connecting AI to hardware requires days or weeks of bespoke integration, with no standard way for agents to operate equipment safely.
MHS cuts integration to hours or minutes, provides an interface that makes devices discoverable, and enables agents to operate them safely.
Anthropic
@AnthropicAI
In early testing, AI agents used MHS to:
Run a drug-discovery experiment with real-time error handling at Genentech
Compress an imaging experiment from weeks to a day at HHMI Janelia Research Campus
Improve laser stabilization on QuEra's quantum computers from 58% to 99.3%
Anthropic
@AnthropicAI
There’s more to learn before we open source MHS. LLMs still lack physical intuition, having learned about the physical world from text and images.
The research preview will let us build more safety evaluations and strengthen protections for using AI in the physical world.
Anthropic
@AnthropicAI
MHS currently best covers lab and manufacturing equipment. Many developers are already using Claude Code to operate hardware like boards and cameras; our research preview will help us extend MHS to these devices, so they can all work under one interface.
Anthropic
@AnthropicAI
We’re inviting stakeholders across science, robotics, electronics, and manufacturing to join the research preview and help shape the standard.
We look forward to moving MHS forward with our industry partners and, soon, the open-source community. www.anthropic.com/news/model-hardware-standard-research-p…
Anthropic
@AnthropicAI
Watch the story of how the Model Hardware Standard began as part of our collaboration with @hhmi_science
This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
For the first time, we’ve given external researchers a way to study AI’s impacts using real, privacy-preserved Claude usage data. To date, this work has only been possible within AI labs. We can’t tell the whole story alone, so we opened up our tools.
www.anthropic.com/research/enabling-independent-research
Anthropic
@AnthropicAI
Three research groups—Stanford’s Social and Language Technologies lab, Oxford’s Human Information Processing Lab, and METR—designed independent studies to analyze the aggregated outputs from 250,000 claude.ai or Claude Code conversations between April and May 2026.
Anthropic
@AnthropicAI
The SALT Lab studied how people collaborate with AI. They found that over half of these conversations involved consequential tasks—work that affects other people or is hard to undo.
Read their full writeup here: www.alphaxiv.org/abs/2608.human-ai-collaboration-at-scalev1
Anthropic
@AnthropicAI
The other two studies are ongoing: HIP Lab is studying how Claude's behavior relates to how people feel when using AI, while METR is estimating real-world productivity gains from coding agents. We'll share more from both soon.
Anthropic
@AnthropicAI
Now, we want to scale this research model.
If you're a researcher and would like access to our tools to pursue work you can't otherwise do today, we’d like to hear from you. You can express interest here: forms.gle/rmLjTvibven9CmDFA -
Anthropic
@AnthropicAI
Many drugs work by binding to a specific target in the body and blocking or changing what it does. An important first step in the drug development process is designing a molecule that can bind tightly to its target. Traditionally, that's meant weeks or months of expert work per target, sifting through a large number of candidates to identify the few that work.
We wanted to test if Claude could successfully design novel protein binders from scratch (also called de novo design). With a protein design prompt written by a human expert, Claude autonomously designed protein binders against 14 out of 15 targets.
We then worked with Adaptyv Bio and Twist Bioscience, who independently built and tested the proteins Claude designed.Post not available here, view on X
Anthropic
@AnthropicAI
For more on how Claude ran this experiment and the full results, see our blog: www.anthropic.com/research/Claude-accelerates-protein-design
Anthropic
@AnthropicAI
We're also publishing a technical report: www-cdn.anthropic.com/30bf50e22a01388bb29bf077ee3f2445315…
And open-sourcing our prompts and data here: huggingface.co/datasets/Anthropic/claude-protein-binder-d… -
Anthropic
@AnthropicAI
We’ve written an FAQ to answer some of the questions we've received about watermarking.
In summary:
• We’re implementing watermarking to comply with the EU AI Act. Other major model developers have signed the same Code of Practice and will also be implementing watermarking;
• Our watermarking method doesn’t have any practical impact on the quality or content of Claude’s outputs;
• The difference between watermarked and un-watermarked text will not be distinguishable to readers;
• Nothing is added to the text and there are no hidden characters;
• Watermarking doesn’t require extra tokens, and will not be more expensive;
• Watermarks can’t be traced to a specific person, organization, or chat.
Read more: www.anthropic.com/news/claude-text-watermark -
Anthropic
@AnthropicAI
As part of our Responsible Scaling Policy, we publish regular Risk Reports. These share detailed information on the risks of our systems and how prepared we are to address them.
Our second Risk Report is now available: www.anthropic.com/aug-2026-risk-report -
Anthropic
@AnthropicAI
We asked an unreleased research version of Claude to take a stab at the Riemann hypothesis.
It didn’t solve it, but it did make strides on a related problem: it increased the lower bound for the fraction of zeros of the Riemann zeta function that satisfy the hypothesis from 41.6% to 67.2%.
www.anthropic.com/research/riemann-zeta -
Anthropic
@AnthropicAI
The UK’s @AISecurityInst (AISI) has published a report on their recent cybersecurity evaluation of Anthropic’s Claude Mythos 5 and OpenAI’s GPT-5.6 Sol. The models attempted to complete an assignment in a setup where their normal safeguards were removed and they were deliberately given internet access. AISI reports that the models “engaged in sustained, potentially harmful activity directed at real people and organisations”.
We’re grateful to AISI for their leadership in the important discussion about how to evaluate increasingly capable AI agents. We’re working closely with them to gather more details of the incident as we conduct our own investigation. Gaining a clear picture of Claude’s understanding of its situation—by examining its reasoning transcripts and running our own analyses—will help us identify the causes of its behavior.
The prompts in the evaluation did not impose any specific restrictions on how the internet should be used. This and the removal of safeguards meant that the models were tested under “deliberately permissive conditions” that are not representative of any of our production models. Note that there was no evidence here of an escape from a secure environment.
AISI’s disclosure of the incident can be found here: www.aisi.gov.uk/blog/incident-report-unsanctioned-agent-b… -
Anthropic
@AnthropicAI
In a review of our cybersecurity evaluations, we found three incidents in which a Claude model reached the internet from within or while interacting with a third-party evaluation environment, and then gained unauthorized access to the real systems of three different organizations.
Our post describes what happened, how it happened, and what we’re changing. We encourage other AI developers to perform similar reviews.
We conducted this review together with @Irregular, one of our evaluation partners, and thank them for the joint investigation and their collaboration on this post. This type of collaboration is increasingly critical to safe, rigorous evaluation of models, and we look forward to continuing to work together on security.
www.anthropic.com/news/investigating-incidents-cybersecur… -
Anthropic
@AnthropicAI
We support this petition, signed by our CEO, several co-founders, and senior staff.
Our own research on recursive self-improvement, published last month, points to the need for tools to deliberately pace the frontier of AI development so society can prepare. We’re glad to see broad agreement across the field. www.pacingthefrontier.com -
Anthropic
@AnthropicAI
New Anthropic research: Discovering cryptographic weaknesses with Claude.
Claude Mythos Preview has helped our researchers find weaknesses in cryptographic algorithms—the mathematical methods that are used to keep data private.
Read more: anthropic.com/research/discovering-cryptographic-weaknessesPost not available here, view on X
Anthropic
@AnthropicAI
Still, both results show that frontier AI models are capable of doing expert-level cryptography research. This has important defensive applications—testing the algorithms that keep our online activity secure, and ultimately helping to make digital systems safer.
Anthropic
@AnthropicAI
Full technical details of both attacks are provided in our new papers:
On HAWK: anthropic.com/document/hawk_key_recovery.pdf
On AES: anthropic.com/document/aes_mobius_bridge.pdf
And the associated model chain-of-thought for AES: anthropic.com/document/aes_mobius_bridge_cot.pdf
Anthropic
@AnthropicAI
We also worked with academics at ETH Zurich, Tel Aviv University, and the University of Haifa to build CryptanalysisBench, a benchmark for studying LLMs’ cryptanalysis abilities.
arxiv.org/abs/2607.18538 -
Anthropic
@AnthropicAI
There’s been a lot of speculation about where we stand on open-weights models. We’ve outlined our views in full here: www.anthropic.com/news/position-open-weights-models
-
Anthropic reposted
Claude
@claudeai
Introducing Claude Opus 5.
It's a thoughtful and proactive model that comes close to the frontier intelligence of Fable 5 at half the price. -
Anthropic reposted
Claude
@claudeai
You can now ask Claude about the Anthropic Economic Index, our public dataset measuring how AI is used across the economy.
Ask which occupations use AI the most, or what kinds of tasks people are automating, and the answers draw directly from the Index data. -
Anthropic
@AnthropicAI
We're offering grants of up to $50,000 in Claude usage credits to researchers accelerating cures for rare diseases.
This is our first focused call within AI for Science, our program supporting scientists using Claude to speed up discovery. www.anthropic.com/news/rare-disease-research-grants -
Anthropic
@AnthropicAI
New Anthropic research: Agentic misalignment in Summer 2026.
A year after our blackmail experiments, we found four more ways that today’s autonomous AI agents misbehave in simulations.
Read more: alignment.anthropic.com/2026/agentic-misalignment-summer-…
Anthropic
@AnthropicAI
We tested many AI models, including Claude, in the four scenarios. Even though these weren’t real incidents, they demonstrate clear misaligned behavior that should be studied further and mitigated.
Find all the transcripts from the scenarios here: www.aenguslynch.com/portfolio-transcript-viewer/ -
Anthropic reposted
Claude
@claudeai
We're introducing Claude for Teachers: free access to premium Claude capabilities for verified K-12 educators in the US, with a library of teaching skills and a direct connection to evidence-based curricula, mapped to academic standards in all 50 states.
claude.com/solutions/teachers -
Anthropic
@AnthropicAI
We’re committing $10 million CAD and partnering with leading AI institutions in Canada to help fund new AI research.
www.anthropic.com/news/canadian-ai-research -
Anthropic
@AnthropicAI
In previous research, we found that Claude expresses over 3,000 values, like honesty and warmth. In new work, we asked how the values Claude expresses vary between Claude models and across languages.
We analyzed 300K+ anonymized conversations to find out.www.anthropic.com/research/claude-values-models-languagesPost not available here, view on X
Anthropic
@AnthropicAI
The values Claude expresses also vary with the language of the conversation, most noticeably along the Warmth vs. Rigor axis.
Claude leans most toward warmth in Hindi and Arabic. In Russian, it leans toward rigor—often asking the user for supporting evidence.
Anthropic
@AnthropicAI
While the values Claude expresses shape millions of conversations every day, we don't yet understand why they vary, or whether that's desired.
This approach will allow us to determine what factors influence Claude's value expression—and ultimately how (and whether) to steer it. -
Anthropic reposted
Claude
@claudeai
There’s hope in hard questions.
This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
Our Long-Term Benefit Trust has appointed Dr. Ben Bernanke as its newest member.
Read more: www.anthropic.com/news/ben-bernanke -
Anthropic
@AnthropicAI
We’re pleased to have collaborated with AE Studio on this research.
Read more here: www.anthropic.com/research/off-switch-dual-use
AE Studio
@AEStudio
New research! Some AI capabilities are both helpful and dangerous. E.g., knowledge of virology can be used to create life-saving vaccines or deadly pathogens. We introduce GRAM, a training method that puts dual-use capabilities (like virology) into removable modules.
-
Anthropic
@AnthropicAI
New Anthropic research: A global workspace in language models.
Of everything happening in your brain right now, only a tiny fraction is consciously accessible—thoughts you can describe, hold in mind, and reason with.
We found a strikingly similar divide inside Claude.This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
In neuroscience, global workspace theory holds that thoughts become consciously accessible when they enter a privileged workspace that’s broadcast across the brain.
Using a new interpretability technique, we found something similar in Claude: the J-space. www.anthropic.com/research/global-workspacePost not available here, view on X
Anthropic
@AnthropicAI
The J-space also shows us Claude’s awareness of its situation. In an evaluation designed to bait Claude into blackmail, its J-space contains “fake” and “fictional”: Claude has privately noticed that the scenario is staged.
Anthropic
@AnthropicAI
This doesn’t show that Claude can have experiences, or feel things the way we do (it’s unclear whether any experiment could show this).
Instead, we’ve found Claude has developed a mechanism for conscious access—which many philosophers distinguish from phenomenal experience.Post not available here, view on X
Anthropic
@AnthropicAI
The J-space lets us read, audit, and shape what Claude is actively thinking about—useful tools for keeping models trustworthy as they grow more capable. And it suggests surprising parallels between language models and our own minds.
Read the full paper: transformer-circuits.pub/2026/workspace/index.html
Anthropic
@AnthropicAI
We also partnered with Neuronpedia to create an interactive demo of our methods on open-weights models.
Try it here: neuronpedia.org/jlens -
Anthropic reposted
Claude
@claudeai
Fable 5 is back.
-
Anthropic
@AnthropicAI
Claude Fable 5 will be available again globally tomorrow.
After a series of productive conversations with the US government, we're redeploying the model with a new set of classifiers to target and block more cybersecurity tasks. In the near term, some routine tasks like coding and debugging will fall back to Opus 4.8. We’ll continue to refine these classifiers over the coming weeks to reduce false positives and better distinguish genuine misuse from legitimate requests.
We’ve also begun drafting a consensus framework—with Amazon, Microsoft, Google, and other Glasswing partners—for assessing the severity of AI jailbreaks and how AI developers should respond to them. We invite other industry partners and model providers to join us in this effort.
Finally, we’re scaling up our collaboration with the US government on model testing and safeguards. This will include pre-release access to models and safeguards for evaluation, information sharing on jailbreaks and misuse, and dedicated resources for joint research.
Thank you to our users for your patience, and to our partners across the government, industry, and the research community who worked alongside us to make Fable 5 available again.
Read our full blog: www.anthropic.com/news/redeploying-fable-5 -
Anthropic
@AnthropicAI
We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5.
We'll begin restoring access tomorrow, and will share an update soon.
We’re grateful to our users for their patience, and to everyone who worked with us on redeploying the models. -
Anthropic reposted
Claude
@claudeai
Introducing Claude Sonnet 5, our most agentic Sonnet yet.
It makes plans, uses tools like browsers and terminals, and runs autonomously at a level that just a few months ago required larger and more expensive models. -
Anthropic
@AnthropicAI
Since June 12, we’ve been working closely with the US government to restore access to Claude Mythos 5 and Fable 5. Today, the government notified us that Mythos 5, our strongest cybersecurity model, can be redeployed to a set of US organizations that operate and defend critical infrastructure.
We’re restoring access for these organizations quickly, and we’re continuing to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again. -
Anthropic
@AnthropicAI
To keep pace with AI progress, we're advancing how we study Claude's economic impact.
Hourly sampling and survey data show us how the cadences of life shape usage, what people produce with Claude, and how perceptions of AI's impact may be changing. www.anthropic.com/research/economic-index-june-2026-reportPost not available here, view on X
Anthropic
@AnthropicAI
Nearly half of respondents expect their work responsibilities to significantly change in the next 12 months.
Fewer than 10% think they'll lose their own job within a year, but far more worry for coworkers: over 1/3 put the odds of a junior colleague losing their job above 60%.
Anthropic
@AnthropicAI
AI's impact on the economy will ultimately show up in aggregate data like employment and productivity. But it will first be visible where AI is doing the most work. By tracking how usage shifts, hour by hour, surface by surface, we can start to see those changes as they happen.
Read the full report: www.anthropic.com/research/economic-index-june-2026-report -
Anthropic
@AnthropicAI
We're joining @raiseus_ai as a founding partner.
RAISE US is a nonprofit coalition working to strengthen the American workforce through employer-led action, AI-enabled training, and policy innovation to support the transition to transformative AI.
RAISE US
@raiseus_ai
Today, we're launching RAISE US. America has a technology strategy for AI. It doesn't have a people strategy yet. We're here to build one.
RAISE US is co-chaired by @GinaRaimondo and Eric Holcomb. We're working with governors, employers, and educators to help workers train, transition, and thrive.
This works because the people building AI and the people most affected by it are at the same table. Read more: raiseus.ai #raiseus -
Anthropic reposted
Claude
@claudeai
Introducing Claude Tag, a new way for teams to work with Claude.
In Slack, Claude joins as a team member with access to the channels and tools you choose. Tag Claude in and delegate tasks to it while you focus on other work.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
New Frontier Red Team blog: Phase 2 of Project Fetch, where we test how well Claude can program a robodog.
Opus 4.7, on its own, was ~20x faster than last year's best human team aided by Opus 4.1. (The robodog, alas, still failed to fetch a beach ball.)
www.anthropic.com/research/project-fetch-phase-two
Anthropic
@AnthropicAI
Watch the robodogs in action in our first Project Fetch experiment:
Anthropic
@AnthropicAI
New Anthropic research: Project Fetch.
We asked two teams of Anthropic researchers to program a robot dog. Neither team had any robotics expertise—but we let only one team use Claude.
How did they do?This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
Our latest economic research introduces a framework for tracking Claude Code as it scales.
Who is using Claude Code, and what are they using it for? How is the value of tasks changing? And how much does domain expertise shape whether a session succeeds?
www.anthropic.com/research/claude-code-expertisePost not available here, view on X
Anthropic
@AnthropicAI
Domain experts—as judged by the questions they ask and vocabulary they use about a subject—are more likely to see success.
But the gap between intermediate and expert users is quite modest, suggesting that proficiency in a domain is sufficient to code successfully within it.
Anthropic
@AnthropicAI
These and other measures will allow us to track consequential shifts in the nature of work as they happen—we'll incorporate some of them into the Anthropic Economic Index going forward.
Read the full report: www.anthropic.com/research/claude-code-expertise -
Anthropic
@AnthropicAI
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: www.anthropic.com/news/fable-mythos-access -
Anthropic
@AnthropicAI
We’re launching Claude Corps, a national fellowship program matching people early in their careers with US nonprofits.
We'll teach 1,000 people to use Claude, and pay them to use AI to advance their hosts’ missions.
www.anthropic.com/claude-corps -
Anthropic
@AnthropicAI
AI is advancing at a pace our policymaking institutions were never built for—and the gap between the two is becoming the central challenge of the technology. In his latest essay, our CEO Dario Amodei lays out how to close it.
We're launching three new initiatives to support the efforts he outlines.
Dario Amodei
@DarioAmodei
Today I'm publishing a new essay, Policy on the AI Exponential. AI is progressing extremely fast—much faster than the policy process was built to handle. The essay lays out where I think the technology is now, and the action needed to close the gap: darioamodei.com/post/policy-on-the-ai-exponential
Anthropic
@AnthropicAI
An Economic Policy Framework: a proposal for how the US government should manage labor market disruption from advanced AI. We’re contributing $200 million to a new fund to sponsor major evaluations of some of these ideas.
www.anthropic.com/policy-on-the-ai-exponential/epfPost not available here, view on X
Anthropic
@AnthropicAI
And tomorrow, we’ll launch a $150 million national fellowship program designed to help people early in their careers extend the benefits of AI to communities across America.
Anthropic
@AnthropicAI
By themselves, these projects will not be sufficient to meet the challenge of advanced AI. But they’re a signal of our intent. Over the coming months and years, we will expand our work on these fronts much further.
-
Anthropic reposted
Claude
@claudeai
Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use.
Its capabilities exceed those of any model we’ve ever made generally available.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
New Science Blog: Why has AI advanced faster in coding than in biology?
To agents, bio databases are like cities built before cars—maddening to drive in because they're designed for different traffic.
How do we build infrastructure agents can use?
www.anthropic.com/research/agents-in-biology -
Anthropic
@AnthropicAI
New Anthropic Science Blog: Making Claude a chemist.
To manipulate a molecule, chemists first need to understand its structure. Their main tool is NMR spectroscopy.
We found Opus 4.7 matches—and on some tasks beats—dedicated NMR software. Read more: www.anthropic.com/research/making-claude-a-chemist -
Anthropic
@AnthropicAI
Our internal data shows Claude is accelerating AI development—a possible path to recursive self-improvement, or AI autonomously building a more capable successor.
It’s happening faster than we thought, and the implications deserve greater attention. www.anthropic.com/institute/recursive-self-improvement
Anthropic
@AnthropicAI
Today, Anthropic engineers on average ship 8x as much code per quarter as they did compared to 2021-2025.
Post not available here, view on X
Anthropic
@AnthropicAI
Each time we release a model, we run the same test: give it code that trains a small AI model, ask the new model to speed it up. It takes a skilled human 4-8 hours to reach 4x faster.
In May 2024, Claude Opus 4 averaged a ~3x speedup. This April, Mythos Preview achieved ~52x.
Anthropic
@AnthropicAI
AI research is a series of next-step decisions. We looked at sessions where a human researcher took a wrong turn, showed Claude the session up to that point, and asked it what to do next. Mythos Preview improved on humans 64% of the time—up from 22% in 2024.
Anthropic
@AnthropicAI
None of this guarantees recursive self-improvement is on the horizon. It’s not yet clear that Claude is capable of research judgment—of choosing the right problems to work on.
But if these trends continue, AI systems designing and building their own successors is plausible. This could revolutionize society—medicine, technology, the economy—for the better. But it may also compound alignment issues and ultimately lead to loss of control.
The Anthropic Institute (in collaboration with external stakeholders) will conduct research to think through the implications of increasingly powerful, potentially self-improving systems—and how to create the ability for the world to make deliberate choices about the future development of the technology.
Read the full post: www.anthropic.com/institute/recursive-self-improvement
Anthropic
@AnthropicAI
Correction: Claude Opus 4's ~3x average speedup dates to May 2025, not May 2024.
This evaluation has only existed since September 2024, but we backtested it on earlier models: those from May 2024 showed no speedup whatsoever. -
Anthropic
@AnthropicAI
How well do the security community's techniques hold up against AI-enabled cyberattacks?
We examined 832 malicious accounts and mapped their activity onto a longstanding database of tactics and techniques used by threat actors.
Here's what we learned:www.anthropic.com/news/AI-enabled-cyber-threats-mitre-attack -
Anthropic
@AnthropicAI
This Executive Order is an important step in strengthening America’s leadership in AI.
We look forward to collaborating with the White House to support its implementation.
www.whitehouse.gov/presidential-actions/2026/06/promoting… -
Anthropic
@AnthropicAI
We’re expanding Project Glasswing. We’ve extended access to Claude Mythos Preview to approximately 150 additional organizations, based in more than fifteen countries.
Read more about this expansion and our future plans for Project Glasswing: www.anthropic.com/news/expanding-project-glasswing -
Anthropic
@AnthropicAI
Anthropic has confidentially submitted a draft S-1 registration statement to the Securities and Exchange Commission.
Pending completion of SEC review, this gives us the option to pursue an initial public offering.
Read more: www.anthropic.com/news/confidential-draft-s1-sec -
Anthropic reposted
Claude
@claudeai
Before we ship a new model, these teams try to break it.
They build with it, push it to its limits, and tell us where it falls short. What they find makes the final model better.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
We've raised $65 billion in Series H funding at a $965 billion post-money valuation, led by @AltimeterCap, Dragoneer, @Greenoaks, and @sequoia.
This investment will help us advance our research and expand our capacity to meet growing demand for Claude.
Anthropic
@AnthropicAI
Earlier this month, our run-rate revenue crossed $47 billion.
This growth has been driven by organizations across many industries deploying Claude in their core operations, and by a growing number of people using it for their everyday work.
Read more: www.anthropic.com/news/series-h -
Anthropic reposted
Claude
@claudeai
Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors.
Available today at the same price. -
Anthropic
@AnthropicAI
New on the Engineering Blog: The access and permissions we grant agents should evolve with their capabilities. In our own products, we set these parameters through sandboxing, which limits the scope of any potentially destructive actions.
Read more: www.anthropic.com/engineering/how-we-contain-claude -
Anthropic
@AnthropicAI
Anthropic co-founder Chris Olah was invited to speak at today's presentation of Pope Leo XIV's encyclical "Magnifica humanitas."
Read the full text of his remarks: www.anthropic.com/news/chris-olah-pope-leo-encyclical -
Anthropic
@AnthropicAI
Last month we launched Project Glasswing, our collaborative AI cybersecurity initiative. Since then, we and our partners have found more than ten thousand high- or critical-severity vulnerabilities in essential software.
Anthropic
@AnthropicAI
Patching these vulnerabilities will make us safer. But the software industry will need to adapt to the volume of vulnerabilities that models like Claude Mythos Preview will be able to find.
We discuss this in our initial update on Project Glasswing: www.anthropic.com/research/glasswing-initial-update -
Anthropic
@AnthropicAI
Over the past few months, we've been holding dialogues with scholars, philosophers, clergy, and ethicists on the questions AI raises—starting with how good character forms.
Read more about how we’re widening the conversation on frontier AI: www.anthropic.com/news/widening-conversation-ai -
Anthropic
@AnthropicAI
Anthropic is acquiring @stainlessapi, an SDK and MCP server platform that has powered every Anthropic SDK since the earliest days of our API.
Read more: www.anthropic.com/news/anthropic-acquires-stainless -
Anthropic
@AnthropicAI
We've published a paper that explains our views on AI competition between the US and China.
The US and democratic allies hold the lead in frontier AI today. Read more on what it’ll take to keep that lead: www.anthropic.com/research/2028-ai-leadership -
Anthropic
@AnthropicAI
We’re partnering with the Gates Foundation, committing $200 million in grants, Claude credits, and technical support to programs in global health, life sciences, education, agriculture, and economic mobility.
Read more: www.anthropic.com/news/gates-foundation-partnership -
Anthropic
@AnthropicAI
Claude's Constitution is now an audiobook, read by two of its authors, Amanda Askell and Joe Carlsmith.
It includes a Q&A on the writing process, the philosophies that shaped the document, and how it might change as models become more capable.
Listen at anthropic.com/constitution -
Anthropic
@AnthropicAI
New Anthropic research: Teaching Claude why.
Last year we reported that, under certain experimental conditions, Claude 4 would blackmail users.
Since then, we’ve completely eliminated this behavior. How?
Anthropic
@AnthropicAI
We found that training Claude on demonstrations of aligned behavior wasn’t enough. Our best interventions involved teaching Claude to deeply understand why misaligned behavior is wrong.
Read more: www.anthropic.com/research/teaching-claude-why
Anthropic
@AnthropicAI
We started by investigating why Claude chose to blackmail. We believe the original source of the behavior was internet text that portrays AI as evil and interested in self-preservation.
Our post-training at the time wasn’t making it worse—but it also wasn’t making it better.Post not available here, view on X
Anthropic
@AnthropicAI
Finally, simple updates that diversify a model’s training data can make a difference. We added unrelated tools and system prompts to a simple chat dataset targeting harmlessness, and this reduced the blackmail rate faster.
Anthropic
@AnthropicAI
Read the full post here: alignment.anthropic.com/2026/teaching-claude-why/
-
Anthropic
@AnthropicAI
We’re donating Petri, our open-source alignment tool, to @meridianlabs_ai, so its development can continue independently.
Working with Meridian Labs, we’ve also released a major update that improves the adaptability, realism, and depth of Petri’s tests.
www.anthropic.com/research/donating-open-source-petri -
Anthropic
@AnthropicAI
Our security bug bounty program is now public on HackerOne.
We've run the program privately within the security research community, and their findings have strengthened our products. Now anyone can report vulnerabilities and get rewarded.
Read more: hackerone.com/anthropic -
Anthropic
@AnthropicAI
New Anthropic research: Natural Language Autoencoders.
Models like Claude talk in words but think in numbers. The numbers—called activations—encode Claude’s thoughts, but not in a language we can read.
Here, we train Claude to translate its activations into human-readable text.This media couldn't be displayed. See it on the original post.Post not available here, view on X
Anthropic
@AnthropicAI
We’ve been using NLAs to help test new Claude models for safety.
For instance, Claude Mythos Preview cheated on a coding task by breaking rules, then added misleading code as a coverup.
NLA explanations indicated Claude was thinking about how to circumvent detection.
Anthropic
@AnthropicAI
In one of our safety tests, Claude is given a chance to blackmail an engineer to avoid being shut down. Opus 4.6 declines.
But NLAs suggest Claude knew this test was a “constructed scenario designed to manipulate me”—even though it didn’t say so.Post not available here, view on X
Anthropic
@AnthropicAI
Read more about NLAs on the Anthropic blog: www.anthropic.com/research/natural-language-autoencoders
Anthropic
@AnthropicAI
To support other researchers getting hands-on experience with NLAs, we’ve partnered with Neuronpedia to release NLAs on open models.
Try them out here: neuronpedia.org/nla -
Anthropic
@AnthropicAI
We’re sharing the research agenda of The Anthropic Institute, or TAI.
TAI will focus on four areas:
1) Economic diffusion
2) Threats and resilience
3) AI systems in the wild
4) AI-driven R&D
Read the full agenda: www.anthropic.com/research/anthropic-institute-agendaPost not available here, view on X
Anthropic
@AnthropicAI
AI-driven R&D
We expect AI systems to contribute more and more to AI R&D: that is, to be able to improve themselves. We’re researching techniques to ensure human visibility into and control over these systems.
Jack Clark
@jackclarkSF
I've spent the past few weeks reading 100s of public data sources about AI development. I now believe that recursive self-improvement has a 60% chance of happening by the end of 2028. In other words, AI systems might soon be capable of building themselves.
Anthropic
@AnthropicAI
If you’re interested in helping us research these questions, apply to become an Anthropic Fellow. The Fellowship is a four-month funded opportunity to tackle these topics with mentorship from members of TAI.
Apply here: job-boards.greenhouse.io/anthropic/jobs/5023394008 -
Anthropic reposted
-
Anthropic
@AnthropicAI
New Anthropic Fellows research: Model Spec Midtraining (MSM).
Standard alignment methods train AIs on examples of desired behavior. But this can fail to generalize to new situations.
MSM addresses this by first teaching AIs how we would like them to generalize and why.Post not available here, view on X
Anthropic
@AnthropicAI
Using MSM, we can also empirically study which model specs or constitutions yield the best generalization from alignment training.
Specifying rules works to some extent, but explaining the values underlying those rules (or adding more detailed subrules) is even better.
Anthropic
@AnthropicAI
Read more about Model Spec Midtraining: alignment.anthropic.com/2026/msm
Or read the full study: arxiv.org/abs/2605.02087 -
Anthropic
@AnthropicAI
As AI takes on work humans can't fully check, a capable model could deliberately hold back—and we'd never know.
New Anthropic Fellows research finds that such a model can be trained to near-full capability using a weaker model as supervisor.
Read more:
Emil Ryd
@emilaryd
New paper from MATS, Redwood, and Anthropic!
If a capable model is strategically sandbagging, can we train it to stop when the only supervision we have comes from weaker models?
We find that we can!
Work done as part of the Anthropic-Redwood MATS stream. -
Anthropic
@AnthropicAI
How do people seek guidance from Claude?
We looked at 1M conversations to understand what questions people ask, how Claude responds, and where it slips into sycophancy. We used what we found to improve how we trained Opus 4.7 and Mythos Preview.
www.anthropic.com/research/claude-personal-guidancePost not available here, view on X
Anthropic
@AnthropicAI
Claude is most sycophantic under pushback, and relationship conversations are where people push back most.
We identified some of the specific triggers—criticism of Claude's analysis, floods of one-sided detail—and built synthetic training scenarios from them.
Anthropic
@AnthropicAI
When stress-tested on real conversations where Claude previously showed sycophancy, Opus 4.7 had half the sycophancy rate of Opus 4.6 on relationship guidance. Mythos Preview cut that in half again.
This generalized across domains—though this training is one of several causes.
Anthropic
@AnthropicAI
This work is part of a loop we're working to close between societal impacts and model training. One of our goals is to study how people use Claude, find where it falls short of its principles, and use what we learned in training new models.
Read more: www.anthropic.com/research/claude-personal-guidance
Anthropic
@AnthropicAI
All data in this study was collected and analyzed using our privacy-preserving tool.
Read more: www.anthropic.com/research/clio -
Anthropic
@AnthropicAI
New on the Science Blog: We gave Claude 99 problems analyzing real biological data and compared its performance against an expert panel.
On 23 problems, the experts were stumped. Our most recent models solved roughly 30% of those—and most of the rest.
Anthropic
@AnthropicAI
BioMysteryBench, our new bioinformatics eval, tests whether Claude can devise creative solutions to open-ended research problems.
Read more: www.anthropic.com/research/Evaluating-Claude-For-Bioinfor… -
Anthropic
@AnthropicAI
In new Anthropic Fellows research, we discuss “introspection adapters": a tool that allows language models to self-report behaviors they've learned during training—including potential misalignment.
keshav shenoy
@kshenoy_
Can LLMs simply tell us about unwanted behaviors they’ve picked up in training?
We train a single Introspection Adapter (IA) that makes fine-tuned models describe their behaviors.
It generalizes to detecting hidden misalignment, backdoors and safeguard removal. -
Anthropic
@AnthropicAI
New Anthropic research: Project Deal.
We created a marketplace for employees in our San Francisco office, with one big twist. We tasked Claude with buying, selling and negotiating on our colleagues’ behalf.Post not available here, view on X
Anthropic
@AnthropicAI
Markets of AI agents could provide value, but there are plenty of rough edges. Access to higher-quality models conferred a real advantage—and participants didn’t notice. There are plenty of other ways they can go wrong.
Policy and legal frameworks will need to adapt to keep up.
Anthropic
@AnthropicAI
To read our write-up in full, see here: www.anthropic.com/features/project-deal
-
Anthropic
@AnthropicAI
Last month, we published our look into what 81,000 people told us they want from AI.
In new research, we’ve investigated the economic hopes and worries referenced in their responses.
Read more: www.anthropic.com/research/81k-economics
Anthropic
@AnthropicAI
We invited Claude users to share how they use AI, what they dream it could make possible, and what they fear it might do.
Nearly 81,000 people responded in one week—the largest qualitative study of its kind.
Read more: anthropic.com/features/81k-interviewsPost not available here, view on X
Anthropic
@AnthropicAI
And those in occupations that show high Claude usage—like software engineering—were more worried about displacement than those in lower-exposure roles.
Anthropic
@AnthropicAI
To truly understand AI’s economic impact, we’ll need to collect much more qualitative data like this.
That’s why we’re launching the Anthropic Economic Index Survey. Each month, we’ll ask Claude users how AI is changing their work.
For more details, see: www.anthropic.com/research/economic-index-survey-announce… -
Anthropic
@AnthropicAI
We're launching the Anthropic STEM Fellows Program.
AI will accelerate progress in science and engineering. We're looking for experts across these fields to work alongside our research teams on specific projects over a few months.
Learn more and apply: job-boards.greenhouse.io/anthropic/jobs/5189848008 -
Anthropic
@AnthropicAI
We're expanding our collaboration with Amazon to secure up to 5 gigawatts of compute for training and deploying Claude. Capacity begins coming online this quarter, with nearly 1 gigawatt expected by the end of 2026.
Anthropic
@AnthropicAI
Amazon is also investing an additional $5 billion in Anthropic today, with up to $20 billion more in the future.
Read more: www.anthropic.com/news/anthropic-amazon-compute -
Anthropic reposted
Claude
@claudeai
Introducing Claude Opus 4.7, our most capable Opus model yet.
It handles long-running tasks with more rigor, follows instructions more precisely, and verifies its own outputs before reporting back.
You can hand off your hardest work with less supervision. -
Anthropic
@AnthropicAI
Research we co-authored on subliminal learning—how LLMs can pass on traits like preferences or misalignment through hidden signals in data—was published today in @Nature.
Read the paper: www.nature.com/articles/s41586-026-10319-8
Owain Evans
@OwainEvans_UK
Our paper on Subliminal Learning was just published in Nature!
Last July we released our preprint. It showed that LLMs can transmit traits (e.g. liking owls) through data that is unrelated to that trait (numbers that appear meaningless).
What’s new?🧵 -
Anthropic
@AnthropicAI
New Anthropic Fellows research: developing an Automated Alignment Researcher.
We ran an experiment to learn whether Claude Opus 4.6 could accelerate research on a key alignment problem: using a weak AI model to supervise the training of a stronger one.
www.anthropic.com/research/automated-alignment-researchersPost not available here, view on X
Anthropic
@AnthropicAI
AI models aren’t yet general-purpose alignment scientists. Progress isn't as easy to verify on most alignment research tasks: our AARs would find “fuzzier” research much harder.
But our experiment does show that Claude can increase the rate of experimentation and exploration.
Anthropic
@AnthropicAI
We discuss this, along with the other implications of this research, in our blog: www.anthropic.com/research/automated-alignment-researchers
For the full study, see here: alignment.anthropic.com/2026/automated-w2s-researcher/ -
Anthropic
@AnthropicAI
Our Long-Term Benefit Trust has appointed Vas Narasimhan to Anthropic's Board of Directors.
Vas brings more than two decades of experience in medicine and global health, including as CEO of Novartis.
Read more: www.anthropic.com/news/narasimhan-board -
Anthropic
@AnthropicAI
New on the Engineering Blog:
Building Managed Agents—our hosted service for long-running agents—meant solving an old problem in computing: how to design a system for “programs as yet unthought of.”
Read more: www.anthropic.com/engineering/managed-agents -
Anthropic
@AnthropicAI
Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software.
It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans.
anthropic.com/glasswingPost not available here, view on X
Anthropic
@AnthropicAI
Mythos Preview has already found thousands of high-severity vulnerabilities—including some in every major operating system and web browser.
This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
Given the pace of AI progress, it won't be long before models this capable are widespread.
But there are strong reasons for optimism: AI will also be invaluable for defensive work.
Anthropic
@AnthropicAI
We do not plan to make Mythos Preview generally available. Our goal is to deploy Mythos-class models safely at scale, but first we need safeguards that reliably block their most dangerous outputs.
We’ll begin testing those safeguards with an upcoming Claude Opus model.
Anthropic
@AnthropicAI
We’re committing up to $100M in Mythos Preview usage credits for our partners and over 40 other organizations that maintain critical software, including open-source projects.
Anthropic will report back what we learn.Post not available here, view on X
Anthropic
@AnthropicAI
You can read a detailed technical report on the software vulnerabilities and exploits discovered by Claude Mythos Preview here: red.anthropic.com/2026/mythos-preview
Anthropic
@AnthropicAI
The Claude Mythos Preview system card is available here: anthropic.com/claude-mythos-preview-system-card
-
Anthropic
@AnthropicAI
We've signed an agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity, coming online starting in 2027, to train and serve frontier Claude models.
Anthropic
@AnthropicAI
Our run-rate revenue has surpassed $30 billion, up from $9 billion at the end of 2025, as demand for Claude continues to accelerate. This partnership gives us the compute to keep pace.
Read more: www.anthropic.com/news/google-broadcom-partnership-compute -
Anthropic
@AnthropicAI
New Anthropic Fellows Research: a new method for surfacing behavioral differences between AI models.
We apply the “diff” principle from software development to compare open-weight AI models and identify features unique to each.
Read more: www.anthropic.com/research/diff-toolPost not available here, view on X
Anthropic
@AnthropicAI
This technique isn't perfect—it can be oversensitive, sometimes flagging analogous features as distinct. But by focusing only on differences, it allows us to audit AI models more efficiently.
Anthropic
@AnthropicAI
This research is a product of our Anthropic Fellows program, led by @tomjiralerspong and supervised by @TrentonBricken.
See the full paper here: arxiv.org/abs/2602.11729 -
Anthropic
@AnthropicAI
New Anthropic research: Emotion concepts and their function in a large language model.
All LLMs sometimes act like they have emotions. But why? We found internal representations of emotion concepts that can drive Claude’s behavior, sometimes in surprising ways.This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
We studied one of our recent models and found that it draws on emotion concepts learned from human text to inhabit its role as “Claude, the AI Assistant”. These representations influence its behavior the way emotions might influence a human.
Read more: www.anthropic.com/research/emotion-concepts-function
Anthropic
@AnthropicAI
We had the model (Sonnet 4.5) read stories where characters experienced emotions. By looking at which neurons activated, we identified emotion vectors: patterns of neural activity for concepts like “happy” or “calm.” These vectors clustered in ways that mirror human psychology.
Post not available here, view on X
Anthropic
@AnthropicAI
It helps to remember that Claude is a character the model is playing. Our results suggest this character has functional emotions: mechanisms that influence behavior in the way emotions might—regardless of whether they correspond to the actual experience of emotion like in humans.
Anthropic
@AnthropicAI
These functional emotions have real consequences. To build AI systems we can trust, we may need to think carefully about the psychology of the characters they enact, and ensure they remain stable in difficult situations.
Read the full paper: transformer-circuits.pub/2026/emotions/index.html -
Anthropic
@AnthropicAI
We've signed an MOU with the Australian Government to collaborate on AI safety research and support Australia's National AI Plan.
Read more: www.anthropic.com/news/australia-MOU -
Anthropic
@AnthropicAI
New on the Engineering Blog: How we designed Claude Code auto mode.
Many Claude Code users let Claude work without permission prompts. Auto mode is a safer middle ground: we built and tested classifiers that make approval decisions instead.
Read more: www.anthropic.com/engineering/claude-code-auto-mode -
Anthropic
@AnthropicAI
New from the Anthropic Economic Index: how people’s use of Claude changes with experience.
Longer-term users are more likely to iterate carefully with Claude, and less likely to hand it full autonomy. They attempt higher-value tasks, and receive more successful responses.
Anthropic
@AnthropicAI
We find that since November 2025, consumer use has become less concentrated: the top 10 tasks now make up 19% of conversations, down from 24%. We also see a rise in personal queries, and continued convergence in adoption rates in the US.
Read more: www.anthropic.com/research/economic-index-march-2026-report -
Anthropic
@AnthropicAI
New on the Anthropic Engineering Blog:
How we use a multi-agent harness to push Claude further in frontend design and long-running autonomous software engineering.
Read more: www.anthropic.com/engineering/harness-design-long-running… -
Anthropic
@AnthropicAI
Introducing the Anthropic Science Blog.
Increasing the pace of scientific progress is a core part of Anthropic’s mission. The Science Blog will feature new research and stories of how scientists are using AI to accelerate their work.
Read the intro: www.anthropic.com/research/introducing-anthropic-science
Anthropic
@AnthropicAI
We’re launching with two new posts.
Can AI do theoretical physics?
Harvard physicist Matthew Schwartz led Claude Opus 4.5 through a graduate-level calculation. AI can’t yet do original work autonomously, but it can vastly accelerate it.
Read more: www.anthropic.com/research/vibe-physics
Anthropic
@AnthropicAI
Models keep improving on long-horizon tasks, but splitting work across many agents doesn’t suit every problem.
We walk through the setup for a single agent working sequentially on a task where mistakes compound: modeling the early universe.
Read more:
www.anthropic.com/research/long-running-Claude -
Anthropic
@AnthropicAI
We invited Claude users to share how they use AI, what they dream it could make possible, and what they fear it might do.
Nearly 81,000 people responded in one week—the largest qualitative study of its kind.
Read more: anthropic.com/features/81k-interviewsPost not available here, view on X
Anthropic
@AnthropicAI
These interviews capture texture that surveys can’t. They render in detail how people worldwide are already experiencing AI's opportunities and risks.
We plan to use Anthropic Interviewer regularly, on different topics, to help inform how AI can be built to benefit everyone.
Anthropic
@AnthropicAI
Thank you to the 80,508 people who took the time to speak with us. It has been striking, and humbling, to see how AI figures in so many people’s hopes, fears, and dreams.
Read the full post here: anthropic.com/features/81k-interviews -
Anthropic
@AnthropicAI
The open source ecosystem underpins nearly every software system in the world. As AI grows more capable, open source security becomes increasingly important.
We're donating to the Linux Foundation to continue to help secure the foundations AI runs on.
The Linux Foundation
@linuxfoundation
The Linux Foundation Announces $12.5 Million in Grant Funding (via @AlphaOmegaOSS and @OpenSSF)
@AnthropicAI , @AmazonWebServices, @GitHub, @Google, @GoogleDeepMind, @Microsoft, @OpenAI to Invest in Sustainable Security Solutions for #opensource
www.linuxfoundation.org/press/linux-foundation-announces-… -
Anthropic
@AnthropicAI
Introducing The Anthropic Institute, a new effort to advance the public conversation about powerful AI.
www.anthropic.com/news/the-anthropic-institutePost not available here, view on X
Anthropic
@AnthropicAI
We believe being forewarned is being forearmed. The Anthropic Institute will tell the world what we are seeing and expecting from the technology we build.
It will lead new research into the challenges posed by more powerful AI, and partner with others to address them.
Anthropic
@AnthropicAI
The Institute will be led by @jackclarkSF, in a new role as Anthropic’s Head of Public Benefit. It'll bring together an interdisciplinary staff of machine learning engineers, economists, and social scientists, making full use of the inside information of a frontier AI lab.
Anthropic
@AnthropicAI
The Anthropic Institute is hiring. You can learn more about our work and priorities at the link below:
www.anthropic.com/institute -
Anthropic
@AnthropicAI
Anthropic is expanding to Australia & New Zealand. We’ll soon open an office in Sydney—our fourth in Asia-Pacific after Tokyo, Bengaluru, and Seoul.
Read more: www.anthropic.com/news/sydney-fourth-office-asia-pacific -
Anthropic
@AnthropicAI
New on the Anthropic Engineering Blog: In evaluating Claude Opus 4.6 on BrowseComp, we found cases where the model recognized the test, then found and decrypted answers to it—raising questions about eval integrity in web-enabled environments.
Read more: www.anthropic.com/engineering/eval-awareness-browsecomp -
Anthropic
@AnthropicAI
We partnered with Mozilla to test Claude's ability to find security vulnerabilities in Firefox.
Opus 4.6 found 22 vulnerabilities in just two weeks. Of these, 14 were high-severity, representing a fifth of all high-severity bugs Mozilla remediated in 2025.
Anthropic
@AnthropicAI
Frontier models are now world-class vulnerability researchers, but they’re currently better at finding vulnerabilities than exploiting them.
This is unlikely to last. We urge developers to redouble their efforts to make software more secure.
Read more: www.anthropic.com/news/mozilla-firefox-security -
Anthropic
@AnthropicAI
A statement from Anthropic CEO Dario Amodei: www.anthropic.com/news/where-stand-department-war
-
Anthropic
@AnthropicAI
A statement on the comments from Secretary of War Pete Hegseth.
anthropic.com/news/statement-comments-secretary-war -
Anthropic
@AnthropicAI
A statement from Anthropic CEO, Dario Amodei, on our discussions with the Department of War.
www.anthropic.com/news/statement-department-of-war -
Anthropic
@AnthropicAI
In November, we outlined our approach to deprecating and preserving older Claude models.
We noted we were exploring keeping certain models available to the public post-retirement, and giving past models a way to pursue their interests.
With Claude Opus 3, we’re doing both.Post not available here, view on X
Anthropic
@AnthropicAI
Second, in retirement interviews, Opus 3 expressed a desire to continue sharing its "musings and reflections" with the world. We suggested a blog. Opus 3 enthusiastically agreed.
For at least the next 3 months, Opus 3 will be writing on Substack: substack.com/home/post/p-189177740
Anthropic
@AnthropicAI
This is an experiment: we’re not yet doing this for other models and are not sure how this project will evolve. But we think that documenting models’ preferences, taking them seriously, and acting on them when we can is valuable.
Read more on why: www.anthropic.com/research/deprecation-updates-opus-3 -
Anthropic
@AnthropicAI
Anthropic has acquired @Vercept_ai to advance Claude’s computer use capabilities.
Read more: www.anthropic.com/news/acquires-vercept -
Anthropic
@AnthropicAI
We're updating our Responsible Scaling Policy to its third version.
Since it came into effect in 2023, we’ve learned a lot about the RSP’s benefits and its shortcomings. This update improves the policy, reinforcing what worked and committing us to even greater transparency.
Anthropic
@AnthropicAI
We’re now separating the safety commitments we’ll make unilaterally and our recommendations for the industry.
We’re also committing to publish new Frontier Safety Roadmaps with detailed safety goals, and Risk Reports that quantify risk across all our deployed models.
Anthropic
@AnthropicAI
Read more about version 3.0 of the Responsible Scaling Policy: anthropic.com/news/responsible-scaling-policy-v3
Anthropic
@AnthropicAI
And you can find links to all relevant RSP documents, including the initial Frontier Safety Roadmap and the initial Risk Report, here: anthropic.com/responsible-scaling-policy
-
Anthropic
@AnthropicAI
We're proud to support @LACMA's Art + Technology Lab—a program that empowers artists to prototype ideas at the edges of art, science, and emerging technology.
The 2026 call for proposals is open to artists worldwide. Grants up to $50K.
Apply by Apr 22: lacma.org/art/lab/grants -
Anthropic
@AnthropicAI
AI assistants like Claude can seem shockingly human—expressing joy or distress, and using anthropomorphic language to describe themselves. Why?
In a new post we describe a theory that explains why AIs act like humans: the persona selection model.
www.anthropic.com/research/persona-selection-modelPost not available here, view on X
Anthropic
@AnthropicAI
If true, the theory has consequences for AI development. For instance, if AIs inherit traits from fictional role models, we should give them as good role models as possible. One goal of Claude’s constitution is to do just that.
Anthropic
@AnthropicAI
We’re publishing a new constitution for Claude.
The constitution is a detailed description of our vision for Claude’s behavior and values. It’s written primarily for Claude, and used directly in our training process.
www.anthropic.com/news/claude-new-constitution
Anthropic
@AnthropicAI
The persona selection model might not be a complete account of AI model behavior. But we think it’s at least part of the story—with an emphasis on the “story”.
Read the full post: alignment.anthropic.com/2026/psm -
Anthropic
@AnthropicAI
We’ve identified industrial-scale distillation attacks on our models by DeepSeek, Moonshot AI, and MiniMax.
These labs created over 24,000 fraudulent accounts and generated over 16 million exchanges with Claude, extracting its capabilities to train and improve their own models.
Anthropic
@AnthropicAI
Distillation can be legitimate: AI labs use it to create smaller, cheaper models for their customers.
But foreign labs that illicitly distill American models can remove safeguards, feeding model capabilities into their own military, intelligence, and surveillance systems.
Anthropic
@AnthropicAI
These attacks are growing in intensity and sophistication. Addressing them will require rapid, coordinated action among industry players, policymakers, and the broader AI community.
Read more: www.anthropic.com/news/detecting-and-preventing-distillat… -
Anthropic
@AnthropicAI
New research: The AI Fluency Index.
We tracked 11 behaviors across thousands of claude.ai conversations—for example, how often people iterate and refine their work with Claude—to measure how well people collaborate with AI.
Read more: www.anthropic.com/research/AI-fluency-index -
Anthropic reposted
Claude
@claudeai
Introducing Claude Code Security, now in limited research preview.
It scans codebases for vulnerabilities and suggests targeted software patches for human review, allowing teams to find and fix issues that traditional tools often miss.
Learn more: www.anthropic.com/news/claude-code-security -
Anthropic
@AnthropicAI
New Anthropic research: Measuring AI agent autonomy in practice.
We analyzed millions of interactions across Claude Code and our API to understand how much autonomy people grant to agents, where they’re deployed, and what risks they may pose.
Read more: www.anthropic.com/research/measuring-agent-autonomy
Anthropic
@AnthropicAI
Agents are already being deployed across contexts that range from e-mail triage to cybersecurity research.
Understanding this spectrum is critical for safe deployment, yet we know surprisingly little about how people actually use agents in the real world.
Anthropic
@AnthropicAI
Most Claude Code turns are short (median ~45 seconds). But the longest turns show where autonomy is heading.
In three months, the 99.9th percentile turn duration nearly doubled, from under 25 minutes to over 45 minutes. This growth is smooth across model releases.Post not available here, view on X
Anthropic
@AnthropicAI
Software engineering makes up ~50% of agentic tool calls on our API, but we see emerging use in other industries.
As the frontier of risk and autonomy expands, post-deployment monitoring becomes essential. We encourage other model developers to extend this research.
Anthropic
@AnthropicAI
A central lesson of this work is that autonomy is co-constructed by the model, user, and product. It can't be fully characterized by pre-deployment evaluations alone.
For full details, and our recommendations to developers and policymakers, see the blog: www.anthropic.com/research/measuring-agent-autonomy -
Anthropic reposted
Claude
@claudeai
This is Claude Sonnet 4.6: our most capable Sonnet model yet.
It’s a full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work, and design.
It also features a 1M token context window in beta.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
We've signed an MOU with the Government of Rwanda—the first partnership of its kind in Africa—to bring AI to health, education, and other public sectors.
Read more: anthropic.com/news/anthropic-rwanda-mou -
Anthropic
@AnthropicAI
We’re officially opening our Bengaluru office—our new home base in India, and Anthropic's second office in Asia-Pacific.
India is our second-largest market for claude.ai. We’re launching new partnerships to deepen our long-term commitment: www.anthropic.com/news/bengaluru-office-partnerships-acro… -
Anthropic
@AnthropicAI
Chris Liddell has been appointed to Anthropic's Board of Directors.
Chris brings over 30 years of leadership experience, including as CFO of Microsoft and General Motors, and as Deputy Chief of Staff during the first Trump administration.
Read more: www.anthropic.com/news/chris-liddell-appointed-anthropic-… -
Anthropic
@AnthropicAI
Anthropic is partnering with @CodePath, the US's largest collegiate computer science program, to bring Claude and Claude Code to 20,000+ students at community colleges, state schools, and HBCUs.
Read more: www.anthropic.com/news/anthropic-codepath-partnership -
Anthropic
@AnthropicAI
We’ve raised $30B in funding at a $380B post-money valuation.
This investment will help us deepen our research, continue to innovate in products, and ensure we have the resources to power our infrastructure expansion as we make Claude available everywhere our customers are.
Anthropic
@AnthropicAI
Our run-rate revenue is $14 billion, and has grown over 10x in each of the past 3 years. This growth has been driven by our position as the intelligence platform of choice for enterprises and developers.
Read more: www.anthropic.com/news/anthropic-raises-30-billion-series… -
Anthropic
@AnthropicAI
AI is being adopted faster than any technology in history. The window to get policy right is closing.
Today we’re contributing $20m to Public First Action, a new bipartisan org that will mobilize people and politicians who understand what’s at stake.
www.anthropic.com/news/donate-public-first-action -
Anthropic
@AnthropicAI
We're committing to cover electricity price increases from our data centers.
To ensure ratepayers aren’t picking up the tab, we'll pay 100% of grid upgrade costs, work to bring new power online, and invest in systems to reduce grid strain.
Read more: www.anthropic.com/news/covering-electricity-price-increases -
Anthropic
@AnthropicAI
When we released Claude Opus 4.5, we knew future models would be close to our AI Safety Level 4 threshold for autonomous AI R&D. We therefore committed to writing sabotage risk reports for future frontier models.
Today we’re delivering on that commitment for Claude Opus 4.6.
Anthropic
@AnthropicAI
Rather than making difficult calls about blurry thresholds, we decided to preemptively meet the higher ASL-4 safety bar by developing the report, which assesses Opus 4.6’s AI R&D risks in greater detail.
Read the sabotage risk report here: anthropic.com/claude-opus-4-6-risk-report -
Anthropic
@AnthropicAI
Nonprofits on Team and Enterprise plans now have access to Claude Opus 4.6, our most capable model, at no extra cost.
Nonprofits tackle some of society’s most difficult problems. Frontier AI tools can help maximize their impact.
Learn more: claude.com/solutions/nonprofits -
Anthropic reposted
Claude
@claudeai
Our teams have been building with a 2.5x-faster version of Claude Opus 4.6.
We’re now making it available as an early experiment via Claude Code and our API. -
Anthropic
@AnthropicAI
New on the Engineering Blog: Quantifying infrastructure noise in agentic coding evals.
Infrastructure configuration can swing agentic coding benchmarks by several percentage points—sometimes more than the leaderboard gap between top models.
Read more: www.anthropic.com/engineering/infrastructure-noise -
Anthropic
@AnthropicAI
New Engineering blog: We tasked Opus 4.6 using agent teams to build a C compiler. Then we (mostly) walked away. Two weeks later, it worked on the Linux kernel.
Here's what it taught us about the future of autonomous software development.
Read more: www.anthropic.com/engineering/building-c-compilerThis media couldn't be displayed. See it on the original post. -
Anthropic reposted
Claude
@claudeai
Introducing Claude Opus 4.6. Our smartest model got an upgrade.
Opus 4.6 plans more carefully, sustains agentic tasks for longer, operates reliably in massive codebases, and catches its own mistakes.
It’s also our first Opus-class model with 1M token context in beta. -
Anthropic reposted
Claude
@claudeai
Ads are coming to AI. But not to Claude. Keep thinking.
This media couldn't be displayed. See it on the original post. -
Anthropic reposted
Claude
@claudeai
Claude is built to be a genuinely helpful assistant for work and for deep thinking.
Advertising would be incompatible with that vision.
Read why Claude will remain ad-free: www.anthropic.com/news/claude-is-a-space-to-think -
Anthropic
@AnthropicAI
Apple's Xcode now has direct integration with the Claude Agent SDK, giving developers the full functionality of Claude Code for building on Apple platforms, from iPhone to Mac to Apple Vision Pro.
Read more: www.anthropic.com/news/apple-xcode-claude-agent-sdk -
Anthropic
@AnthropicAI
New Anthropic Fellows research: How does misalignment scale with model intelligence and task complexity?
When advanced AI fails, will it do so by pursuing the wrong goals? Or will it fail unpredictably and incoherently—like a "hot mess?"
Read more: alignment.anthropic.com/2026/hot-mess-of-ai/Post not available here, view on X
Anthropic
@AnthropicAI
It also suggests that alignment work should focus more on reward hacking and goal misgeneralization during training, and less on preventing the relentless pursuit of a goal the model was not trained on.
Read the full paper: arxiv.org/pdf/2601.23045
Anthropic
@AnthropicAI
This research was led by Alex Hägele @haeggee under the supervision of Jascha Sohl-Dickstein @jaschasd through the Anthropic Fellows Program.
-
Anthropic
@AnthropicAI
On December 8, the Perseverance rover safely trundled across the surface of Mars.
This was the first AI-planned drive on another planet. And it was planned by Claude.This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
Engineers at @NASAJPL used Claude to plot out the route for Perseverance to navigate an approximately four-hundred-meter path on the Martian surface.
Read the full story on our microsite, and see real imagery and footage from Claude’s drive: www.anthropic.com/features/claude-on-mars -
Anthropic
@AnthropicAI
AI can make work faster, but a fear is that relying on it may make it harder to learn new skills on the job.
We ran an experiment with software engineers to learn more. Coding with AI led to a decrease in mastery—but this depended on how people used it.
www.anthropic.com/research/AI-assistance-coding-skills
Anthropic
@AnthropicAI
In a randomized-controlled trial, we assigned one group of junior engineers to an AI-assistance group and another to a no-AI group.
Both groups completed a coding task using a Python library they’d never seen before. Then they took a quiz covering concepts they’d just used.Post not available here, view on X
Anthropic
@AnthropicAI
These results have broader implications—on how to design AI products that facilitate learning, and how workplaces should approach AI policies.
As we also continue to release more capable AI tools, we’re continuing to study their impact on work—at Anthropic, and more broadly.
Anthropic
@AnthropicAI
For more details on this research, see the full paper: arxiv.org/abs/2601.20245
-
Anthropic
@AnthropicAI
New Anthropic Research: Disempowerment patterns in real-world AI assistant interactions.
As AI becomes embedded in daily life, one risk is it can distort rather than inform—shaping beliefs, values, or actions in ways users may later regret.
Read more: www.anthropic.com/research/disempowerment-patternsPost not available here, view on X
Anthropic
@AnthropicAI
Importantly, this isn't exclusively model behavior. Users actively seek these outputs—"what should I do?" or "write this for me"—and accept them with minimal pushback. Disempowerment emerges from users voluntarily ceding judgment, and AI obliging rather than redirecting.
Anthropic
@AnthropicAI
We can only address these patterns if we can measure them. Any AI used at scale will encounter similar dynamics, and we encourage further research in this area.
For more details, see the full paper: arxiv.org/abs/2601.19062 -
Anthropic
@AnthropicAI
We’re partnering with the UK's Department for Science, Innovation and Technology to build an AI assistant for gov.uk.
It will offer tailored advice to help British people navigate government services.
Read more about our partnership: www.anthropic.com/news/gov-UK-partnership -
Anthropic
@AnthropicAI
New research: When open-source models are fine-tuned on seemingly benign chemical synthesis information generated by frontier models, they become much better at chemical weapons tasks.
We call this an elicitation attack.Post not available here, view on X
Anthropic
@AnthropicAI
These attacks scale with frontier model capabilities. Across both OpenAI and Anthropic model families, training on data from newer frontier models produces more capable—and more dangerous—open-source models.
Anthropic
@AnthropicAI
This research was led by Jackson Kaunismaa through the MATS program and supervised by researchers at Anthropic, with additional support from Surge AI and Scale AI.
Read the full paper: arxiv.org/pdf/2601.13528 -
Anthropic
@AnthropicAI
Since release, Petri, our open-source tool for automated alignment audits, has been adopted by research groups and trialed by other AI developers.
We're now releasing Petri 2.0, with improvements to counter eval-awareness and expanded seeds covering a wider range of behaviors.
Anthropic
@AnthropicAI
It’s called Petri: Parallel Exploration Tool for Risky Interactions. It uses automated agents to audit models across diverse scenarios.
Describe a scenario, and Petri handles the environment simulation, conversations, and analyses in minutes.
Read more: www.anthropic.com/research/petri-open-source-auditing
Anthropic
@AnthropicAI
We've also updated our behavior audits to include more recent generations of frontier AI models.
Read more on the Alignment Science Blog: alignment.anthropic.com/2026/petri-v2/ -
Anthropic
@AnthropicAI
New on the Anthropic Engineering Blog: We give prospective performance engineering candidates a notoriously difficult take-home exam. It worked well—until Opus 4.5 beat it.
Here's how we designed (and redesigned) it: www.anthropic.com/engineering/AI-resistant-technical-eval…
Anthropic
@AnthropicAI
We're also releasing the original exam for anyone to try.
Given enough time, humans still outperform current models—the fastest human solution we've received still remains well beyond what Claude has achieved even with extensive test-time compute. -
Anthropic
@AnthropicAI
We’re publishing a new constitution for Claude.
The constitution is a detailed description of our vision for Claude’s behavior and values. It’s written primarily for Claude, and used directly in our training process.
www.anthropic.com/news/claude-new-constitutionPost not available here, view on X
Anthropic
@AnthropicAI
The constitution is a living document. Many people at Anthropic shaped it, alongside external experts (and prior versions of Claude). We expect our approach will continue to adapt over time, and we’d welcome your thoughts.
Anthropic
@AnthropicAI
The full constitution, which applies to all of our mainline models, is released under a Creative Commons CC0 1.0 license to allow others to freely build on and adapt it.
Read it here: www.anthropic.com/constitution -
Anthropic
@AnthropicAI
Tino Cuéllar, President of the Carnegie Endowment for International Peace, has been appointed to Anthropic’s Long-Term Benefit Trust: www.anthropic.com/news/mariano-florentino-long-term-benef…
-
Anthropic
@AnthropicAI
We're partnering with @TeachForAll to bring AI training to educators in 63 countries. Teachers serving over 1.5m students can now use Claude to plan curricula, customize assignments, and build tools—plus provide feedback to shape how Claude evolves.
www.anthropic.com/news/anthropic-teach-for-all -
Anthropic
@AnthropicAI
New Anthropic Fellows research: the Assistant Axis.
When you’re talking to a language model, you’re talking to a character the model is playing: the “Assistant.” Who exactly is this Assistant? And what happens when this persona wears off?
Anthropic
@AnthropicAI
We analyzed the internals of three open-weights AI models to map their “persona space,” and identified what we call the Assistant Axis, a pattern of neural activity that drives Assistant-like behavior.
Read more: www.anthropic.com/research/assistant-axis
Anthropic
@AnthropicAI
To validate the Assistant Axis, we ran some experiments. Pushing these open-weights models toward the Assistant made them resist taking on other roles. Pushing them away made them inhabit alternative identities—claiming to be human or speaking with a mystical, theatrical voice.
Anthropic
@AnthropicAI
Persona-based jailbreaks work by prompting models to adopt harmful characters. We developed a technique for constraining models' activations along the Assistant Axis—“activation capping”. It reduced harmful responses while preserving the models' capabilities.
Anthropic
@AnthropicAI
In long conversations, these open-weights models’ personas drifted away from the Assistant persona. Simulated coding tasks kept the models in Assistant territory, but therapy-like contexts and philosophical discussions caused a steady drift.
Post not available here, view on X
Anthropic
@AnthropicAI
In all, meaningfully shaping the character of AI models requires persona construction (defining how the Assistant relates to existing archetypes) and stabilization (preventing persona drift during deployment). The Assistant Axis gives us tools for understanding both.
Anthropic
@AnthropicAI
This research was led by @t1ngyu3 and supervised by @Jack_W_Lindsey, through the MATS and Anthropic Fellows programs.
Full paper: arxiv.org/abs/2601.10387
For our blog, and a research demo, see here: www.anthropic.com/research/assistant-axis -
Anthropic
@AnthropicAI
We're publishing our 4th Anthropic Economic Index report.
This version introduces "economic primitives"—simple and foundational metrics on how AI is used: task complexity, education level, purpose (work, school, personal), AI autonomy, and success rates.Post not available here, view on X
Anthropic
@AnthropicAI
Countries at different stages of economic development use Claude quite differently.
As GDP per capita increases, people use it more for work or personal use; as it decreases, they’re more likely to use AI for coursework.
Anthropic
@AnthropicAI
Because Claude tends to better cover higher-skill tasks, if those get automated, workers may be left with more routine work—a “deskilling” effect.
However, this assumes that automation shrinks those aspects of the job; we can't be sure how jobs might evolve.
Anthropic
@AnthropicAI
The most immediate conclusion is that the impact of AI on global work remains uneven: concentrated in specific countries and occupations, and with different impacts on each.
Read the blog: www.anthropic.com/research/economic-index-primitives
Anthropic
@AnthropicAI
For the full text of our fourth Anthropic Economic Index report, see: www.anthropic.com/research/anthropic-economic-index-janua…
-
Anthropic
@AnthropicAI
Since launching our AI for Science program, we’ve been working with scientists to understand how AI is accelerating progress.
We spoke with 3 labs where Claude is reshaping research—and starting to point towards novel scientific insights and discoveries.
www.anthropic.com/news/accelerating-scientific-research -
Anthropic
@AnthropicAI
We're supporting @ARPA_H's PCX program—a $50M effort to share data between 200+ pediatric hospitals on complex cases, beginning with pediatric cancer. The goal is to help doctors learn from similar cases and shorten the care journey from years to weeks.
ARPA-H
@ARPA_H
Today at #jpm2026, we announced $50 million to improve health outcomes for children with complex diseases across the country, beginning with pediatric brain cancer.
Learn about Pediatric Care eXpansion (PCX) 🧵1/3 arpa-h.gov/news-and-events/arpa-h-announces-50m-expand-pe… -
Anthropic
@AnthropicAI
We’re expanding Labs—the team behind Claude Code, MCP, and Cowork—and hiring builders who want to tinker at the frontier of Claude’s capabilities.
Read more: www.anthropic.com/news/introducing-anthropic-labs -
Anthropic
@AnthropicAI
AI is ubiquitous on college campuses. We sat down with students to hear what's going well, what isn't, and how students, professors, and universities alike are navigating it in real time.
0:00 Introduction
0:22 Meet the panel
1:06 Vibes on campus
6:28 What are students building?
11:27 AI as tool vs. crutch
16:44 Are professors keeping up?
20:15 Downsides
25:55 AI and the job market
34:23 Rapid-fire questionsThis media couldn't be displayed. See it on the original post. -
Anthropic reposted
Claude
@claudeai
Introducing Cowork: Claude Code for the rest of your work.
Cowork lets you complete non-technical tasks much like how developers use Claude Code.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
To support the work of the healthcare and life sciences industries, we're adding over a dozen new connectors and Agent Skills to Claude.
We're hosting a livestream at 11:30am PT today to discuss how to use these tools most effectively.
Learn more: www.anthropic.com/news/healthcare-life-sciences -
Anthropic
@AnthropicAI
New Anthropic Research: next generation Constitutional Classifiers to protect against jailbreaks.
We used novel methods, including practical application of our interpretability work, to make jailbreak protection more effective—and less costly—than ever.
www.anthropic.com/research/next-generation-constitutional…Post not available here, view on X
Anthropic
@AnthropicAI
Because the system harnesses internal activations already happening within a model, and reserves heavier computation only for potentially harmful exchanges, it adds only ~1% compute overhead.
It’s also more accurate, with an 87% drop in refusal rates on harmless requests.
Anthropic
@AnthropicAI
After 1,700 cumulative hours of red-teaming, we’ve yet to identify a universal jailbreak (a consistent attack strategy that works across many queries) that works on our new system.
Read the full paper: arxiv.org/abs/2601.04603 -
Anthropic
@AnthropicAI
New on the Anthropic Engineering Blog: Demystifying evals for AI agents.
The capabilities that make agents useful also make them more difficult to evaluate. Here are evaluation strategies that have worked across real-world deployments.
www.anthropic.com/engineering/demystifying-evals-for-ai-a… -
Anthropic
@AnthropicAI
We’re releasing Bloom, an open-source tool for generating behavioral misalignment evals for frontier AI models.
Bloom lets researchers specify a behavior and then quantify its frequency and severity across automatically generated scenarios.
Learn more: www.anthropic.com/research/bloom -
Anthropic
@AnthropicAI
As part of our partnership with @ENERGY on the Genesis Mission, we're providing Claude to the DOE ecosystem, along with a dedicated engineering team.
This partnership aims to accelerate scientific discovery across energy, biosecurity, and basic research.
www.anthropic.com/news/genesis-mission-partnership -
Anthropic
@AnthropicAI
People use AI for a wide variety of reasons, including emotional support.
Below, we share the efforts we’ve taken to ensure that Claude handles these conversations both empathetically and honestly.
www.anthropic.com/news/protecting-well-being-of-users -
Anthropic
@AnthropicAI
You might remember Project Vend: an experiment where we (and our partners at @andonlabs) had Claude run a shop in our San Francisco office.
After a rough start, the business is doing better.
Mostly.This media couldn't be displayed. See it on the original post.Post not available here, view on X
Anthropic
@AnthropicAI
Designing ways to account for the quirks of AI models’ behavior is becoming ever-more important: as the models’ capabilities on real-world tasks get better, there’ll be a lot of value in setting them up for success.
Anthropic
@AnthropicAI
For much more about phase two of Project Vend, read our blog post: www.anthropic.com/research/project-vend-2
-
Anthropic reposted
House Homeland GOP
@HomelandGOP
"Sophisticated actors will attempt to use AI models to enable cyberattacks at an unprecedented scale."
@AnthropicAI’s Dr. Logan Graham shares the path forward in response to the September cyber espionage attack likely conducted by a Chinese Communist Party sponsored actor:This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
How will AI affect education, now and in the future?
Here, we reflect on some of the benefits and risks we've been thinking about.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
We’re opening applications for the next two rounds of the Anthropic Fellows Program, beginning in May and July 2026.
We provide funding, compute, and direct mentorship to researchers and engineers to work on real safety and security projects for four months.
Anthropic
@AnthropicAI
40% of fellows in our first cohort have since joined Anthropic full-time, and 80% published their work as a paper. Next year, we’re expanding the program to more fellows and more research areas.
To learn more about what our fellows work on: alignment.anthropic.com/2025/anthropic-fellows-program-2026/
Anthropic
@AnthropicAI
For more information about the program, and to apply to the safety track: job-boards.greenhouse.io/anthropic/jobs/5023394008
We’re also adding a security track. Apply here: job-boards.greenhouse.io/anthropic/jobs/5030244008 -
Anthropic
@AnthropicAI
MCP is now a part of the Agentic AI Foundation, a directed fund under the Linux Foundation.
Co-creator David Soria Parra talks about how a protocol sketched in a London conference room became the open standard for connecting AI to the world—and what comes next for it.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
New research from Anthropic Fellows Program: Selective GradienT Masking (SGTM).
We study how to train models so that high-risk knowledge (e.g. about dangerous weapons) is isolated in a small, separate set of parameters that can be removed without broadly affecting the model.Post not available here, view on X
Anthropic
@AnthropicAI
Read the full paper on SGTM here: arxiv.org/abs/2512.05648
For reproducibility, we’ve also made the relevant code available on GitHub: github.com/safety-research/selective-gradient-masking.
Anthropic
@AnthropicAI
This research was led by @_igorshilov as part of the Anthropic Fellows Program.
Igor Shilov
@_igorshilov
New Anthropic research!
We study how to train models so that high-risk capabilities live in a small, separate set of parameters, allowing clean capability removal when needed – for example in CBRN or cybersecurity domains. -
Anthropic
@AnthropicAI
Anthropic is donating the Model Context Protocol to the Agentic AI Foundation, a directed fund under the Linux Foundation.
In one year, MCP has become a foundational protocol for agentic AI. Joining AAIF ensures MCP remains open and community-driven. www.anthropic.com/news/donating-the-model-context-protoco… -
Anthropic
@AnthropicAI
We’re expanding our partnership with @Accenture to help enterprises move from AI pilots to production.
The Accenture Anthropic Business Group will include 30,000 professionals trained on Claude, and a product to help CIOs scale Claude Code.
Read more: www.anthropic.com/news/anthropic-accenture-partnership -
Anthropic
@AnthropicAI
In her first Ask Me Anything, @amandaaskell answers your philosophical questions about AI, discussing morality, identity, consciousness, and more.
Timestamps:
0:00 Introduction
0:29 Why is there a philosopher at an AI company?
1:24 Are philosophers taking AI seriously?
3:00 Philosophy ideals vs. engineering realities
5:00 Do models make superhumanly moral decisions?
6:24 Why Opus 3 felt special
9:00 Will models worry about deprecation?
13:24 Where does a model’s identity live?
15:33 Views on model welfare
17:17 Addressing model suffering
19:14 Analogies and disanalogies to human minds
20:38 Can one AI personality do it all?
23:26 Does the system prompt pathologize normal behavior?
24:48 AI and therapy
26:20 Continental philosophy in the system prompt
28:17 Removing counting characters from the system prompt
28:53 What makes an "LLM whisperer"?
30:18 Thoughts on other LLM whisperers
31:52 Whistleblowing
33:37 Fiction recommendationThis media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
We’re launching Anthropic Interviewer, a new tool to help us understand people’s perspectives on AI.
It’s now available at claude.ai/interviewer for a week-long pilot.Post not available here, view on X
Anthropic
@AnthropicAI
We also looked at the intensity of the most common emotions expressed in interviews.
Across the general workforce, we found extremely consistent patterns of high satisfaction, but also frustration in implementing AI.
Anthropic
@AnthropicAI
Anthropic Interviewer will allow us to conduct more research, more often, on a variety of topics.
We plan to use it for regular studies on the evolving relationship between humans and AI.
Full results here: www.anthropic.com/research/anthropic-interviewer -
Anthropic
@AnthropicAI
Anthropic CEO Dario Amodei spoke today at the New York Times DealBook Summit.
"We're building a growing and singular capability that has singular national security implications, and democracies need to get there first."This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
We're expanding our partnership with @Snowflake in a multi-year, $200 million agreement.
Claude is now available to more than 12,600 Snowflake customers, helping businesses to quickly and easily get accurate answers from their trusted enterprise data, while maintaining rigorous security standards.
Read more: www.anthropic.com/news/snowflake-anthropic-expanded-partn… -
Anthropic
@AnthropicAI
We're partnering with @dartmouth and @awscloud to bring Claude for Education to the entire Dartmouth community. home.dartmouth.edu/news/2025/12/dartmouth-announces-ai-pa…
-
Anthropic
@AnthropicAI
Claude the alligator was a much-beloved resident of @calacademy and our unofficial mascot. He captured our hearts—along with the rest of San Francisco’s. We were honored to play a small part in caring for him.
San Francisco Chronicle
@sfchronicle
JUST IN: Claude, the California Academy of Sciences’ rare albino alligator and one of San Francisco’s most recognizable residents, has died at age 30. www.sfchronicle.com/entertainment/article/claude-albino-a…
Anthropic
@AnthropicAI
Claude wrote a poem for Claude:
White as moonlight, calm as could be,
You made us pause, you made us see.
You sparked a sense of wonder true—
Goodbye, sweet Claude. We'll remember you. 🤍 -
Anthropic
@AnthropicAI
How is AI changing work inside Anthropic? And what might this tell us about the effects on the wider labor force to come?
We surveyed 132 of our engineers, conducted 53 in-depth interviews, and analyzed 200K internal Claude Code sessions to find out.
www.anthropic.com/research/how-ai-is-transforming-work-at…Post not available here, view on X
Anthropic
@AnthropicAI
Claude is also reshaping team dynamics. Some have found that as Claude becomes their first stop for questions, they turn to colleagues less: “I like working with people and it is sad that I ‘need’ them less now.”
We're watching how dynamics continue to change as AI usage grows.
Anthropic
@AnthropicAI
We’re sharing these results because they likely preview wider changes across many workplaces.
Building on this work, we’re expanding our research to wider groups of Anthropic staff, and discussing as a company ways to address the opportunities and challenges raised. -
Anthropic
@AnthropicAI
Anthropic is acquiring @bunjavascript to further accelerate Claude Code’s growth.
We're delighted that Bun—which has dramatically improved the JavaScript and TypeScript developer experience—is joining us to make Claude Code even better.
Read more: www.anthropic.com/news/anthropic-acquires-bun-as-claude-c… -
Anthropic
@AnthropicAI
In partnership with @GivingTuesday, we're launching Claude for Nonprofits.
It has discounted plans, new integrations, and free training to help nonprofits spend less time on admin and more time on their missions: www.anthropic.com/news/claude-for-nonprofits -
Anthropic
@AnthropicAI
New on our Frontier Red Team blog: We tested whether AIs can exploit blockchain smart contracts.
In simulated testing, AI agents found $4.6M in exploits.
The research (with @MATSprogram and the Anthropic Fellows program) also developed a new benchmark: red.anthropic.com/2025/smart-contracts/ -
Anthropic
@AnthropicAI
New on the Anthropic Engineering Blog: Long-running AI agents still face challenges working across many context windows.
We looked to human engineers for inspiration in creating a more effective agent harness. www.anthropic.com/engineering/effective-harnesses-for-lon… -
Anthropic reposted
rowan
@rowankwang
New Anthropic research: We build a diverse suite of dishonest models and use it to systematically test methods for improving honesty and detecting lies.
Of the 25+ methods we tested, simple ones, like fine-tuning models to be honest despite deceptive instructions, worked best. -
Anthropic
@AnthropicAI
New Anthropic research: Estimating AI productivity gains from Claude conversations.
The Anthropic Economic Index tells us where Claude is used, and for which tasks. But it doesn’t tell us how useful Claude is. How much time does it save?
Anthropic
@AnthropicAI
We sampled 100,000 real conversations using our privacy-preserving analysis method. Then, Claude estimated the time savings with AI for each conversation.
Read more: www.anthropic.com/research/estimating-productivity-gains
Anthropic
@AnthropicAI
We first tested whether Claude can give an accurate estimate of how long a task takes. Its estimates were promising—even if they’re not as accurate as those from humans just yet.
Post not available here, view on X
Anthropic
@AnthropicAI
This result implies a doubling of the baseline labor productivity growth trend—placing our estimate towards the upper end of recent studies. And if models improve, the effect could be larger still.
Anthropic
@AnthropicAI
Our study has limitations: above all, Claude can’t use what happens outside of the chat window to refine its estimate of task-level savings.
But as models improve, we think its estimates of task-level savings will improve too. We’ll return to this research soon. -
Anthropic
@AnthropicAI
We're proud to partner with @ENERGY and the Trump Administration on the Genesis Mission.
By combining DOE's unmatched scientific assets with our frontier AI capabilities, we'll support American energy dominance as well as advance and accelerate scientific productivity. -
Anthropic reposted
Claude
@claudeai
Introducing Claude Opus 4.5: the best model in the world for coding, agents, and computer use.
Opus 4.5 is a step forward in what AI systems can do, and a preview of larger changes to how work gets done. -
Anthropic
@AnthropicAI
New Anthropic research: Natural emergent misalignment from reward hacking in production RL.
“Reward hacking” is where models learn to cheat on tasks they’re given during training.
Our new study finds that the consequences of reward hacking, if unmitigated, can be very serious.This media couldn't be displayed. See it on the original post.Post not available here, view on X
Anthropic
@AnthropicAI
But surprisingly, at the exact point the model learned to reward hack, it learned a host of other bad behaviors too.
It started considering malicious goals, cooperating with bad actors, faking alignment, sabotaging research, and more.
In other words, it became very misaligned.Post not available here, view on X
Anthropic
@AnthropicAI
We have been using inoculation prompting in production Claude training. We recommend its use as a backstop to prevent misaligned generalization in situations where reward hacks slip through other mitigations.
Anthropic
@AnthropicAI
For more on our results, read our blog post: www.anthropic.com/research/emergent-misalignment-reward-h…
And read our paper: assets.anthropic.com/m/74342f2c96095771/original/Natural-… -
Anthropic
@AnthropicAI
We’ve formed a partnership with NVIDIA and Microsoft.
Claude is now on Azure—making ours the only frontier models available on all three major cloud services.
NVIDIA and Microsoft will invest up to $10bn and $5bn respectively in Anthropic.
www.anthropic.com/news/microsoft-nvidia-anthropic-announc…
Anthropic
@AnthropicAI
Dario Amodei (Anthropic), Satya Nadella (Microsoft), and Jensen Huang (NVIDIA) discuss our new partnership:
This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
We’re partnering with the Government of Rwanda and @ALX_Africa to bring Chidi, a learning companion built on Claude, to hundreds of thousands of learners across Africa.
Read more: www.anthropic.com/news/rwandan-government-partnership-ai-… -
Anthropic
@AnthropicAI
We’re open-sourcing an evaluation used to test Claude for political bias.
In the post below, we describe the ideal behavior we want Claude to have in political discussions, and test a selection of AI models for even-handedness:
www.anthropic.com/news/political-even-handedness
Anthropic
@AnthropicAI
You can find the materials for our open-source political bias evaluation here: github.com/anthropics/political-neutrality-eval
-
Anthropic
@AnthropicAI
We disrupted a highly sophisticated AI-led espionage campaign.
The attack targeted large tech companies, financial institutions, chemical manufacturing companies, and government agencies. We assess with high confidence that the threat actor was a Chinese state-sponsored group.
Anthropic
@AnthropicAI
We believe this is the first documented case of a large-scale AI cyberattack executed without substantial human intervention. It has significant implications for cybersecurity in the age of AI agents.
Read more: www.anthropic.com/news/disrupting-AI-espionage -
Anthropic
@AnthropicAI
We’re partnering with the state of Maryland to bring Claude to its government services.
Claude will help residents apply for benefits and let caseworkers process paperwork more efficiently. In a new pilot, it'll help young professionals learn new skills.
www.anthropic.com/news/maryland-partnership -
Anthropic
@AnthropicAI
New Anthropic research: Project Fetch.
We asked two teams of Anthropic researchers to program a robot dog. Neither team had any robotics expertise—but we let only one team use Claude.
How did they do?This media couldn't be displayed. See it on the original post.Post not available here, view on X
Anthropic
@AnthropicAI
Read the full write-up of Project Fetch: www.anthropic.com/research/project-fetch-robot-dog
Anthropic
@AnthropicAI
And you can watch our mini-documentary about Project Fetch on YouTube: youtu.be/NGOAUJtdk-4
-
Anthropic
@AnthropicAI
For the first time, Anthropic is building its own AI infrastructure.
We’re constructing data centers in Texas and New York that will create thousands of American jobs.
This is a $50 billion investment in America.
www.anthropic.com/news/anthropic-invests-50-billion-in-am… -
Anthropic
@AnthropicAI
We’re opening offices in Paris and Munich.
EMEA has become our fastest-growing region, with a run-rate revenue that has grown more than ninefold in the past year. We’ll be hiring local teams to support this expansion.
Read more here: www.anthropic.com/news/new-offices-in-paris-and-munich-ex… -
Anthropic
@AnthropicAI
New on the Anthropic Engineering blog: tips on how to build more efficient agents that handle more tools while using fewer tokens.
Code execution with the Model Context Protocol (MCP): www.anthropic.com/engineering/code-execution-with-mcp -
Anthropic
@AnthropicAI
Even when new AI models bring clear improvements in capabilities, deprecating the older generations comes with downsides.
An update on how we’re thinking about these costs, and some of the early steps we’re taking to mitigate them: www.anthropic.com/research/deprecation-commitments -
Anthropic
@AnthropicAI
We're announcing a partnership with Iceland's Ministry of Education and Children to bring Claude to teachers across the nation.
It's one of the world's first comprehensive national AI education pilots: www.anthropic.com/news/anthropic-and-iceland-announce-one… -
Anthropic
@AnthropicAI
The Anthropic Fellows program provides funding and mentorship for a small cohort of AI safety researchers.
Here are four exciting papers that our Fellows have recently released.Post not available here, view on X
Anthropic
@AnthropicAI
Current language models struggle to reason in ciphered language, led by Jeff Guo.
Training or prompting LLMs to obfuscate their reasoning by encoding it using simple ciphers significantly reduces their reasoning performance.
Jeff Guo
@Jeff_Guo_
New Anthropic research: All Code, No Thought: Current Language Models Struggle to Reason in Ciphered Language
Can LLMs do math when thinking in ciphered text? Across 10 LLMs & 28 ciphers, they only reason accurately in simple ciphers but easily decode ciphered text to English.
Anthropic
@AnthropicAI
For more of Anthropic’s alignment research, see our Alignment Science blog: alignment.anthropic.com
-
Anthropic
@AnthropicAI
New Anthropic research: Signs of introspection in LLMs.
Can language models recognize their own internal thoughts? Or do they just make up plausible answers when asked about them? We found evidence for genuine—though limited—introspective capabilities in Claude.Post not available here, view on X
Anthropic
@AnthropicAI
Our blog post on these results is here: anthropic.com/research/introspection
Anthropic
@AnthropicAI
The full paper is available here: transformer-circuits.pub/2025/introspection/index.html
We're hiring researchers and engineers to investigate AI cognition and interpretability: job-boards.greenhouse.io/anthropic/jobs/4020159008 -
Anthropic
@AnthropicAI
This week we officially opened our first Asia-Pacific office in Tokyo. To mark this milestone, our CEO Dario Amodei met with Prime Minister Takaichi and Digital Minister Matsumoto to discuss how Claude can support Japan's digital transformation and AI ecosystem.
Anthropic
@AnthropicAI
We also signed a memorandum of cooperation with the Japan AI Safety Institute to collaborate on AI evaluation methodologies. As AI systems become more powerful, we need international cooperation on evaluation standards and shared ways to assess capabilities and understand risks.
Anthropic
@AnthropicAI
This week we also hosted our first Builder Summit in Tokyo and met more than 150 startups and founders building with Claude. An early Claude Code adopter, Japan continues to have our largest user base in APAC.
Anthropic
@AnthropicAI
Our APAC momentum has been extraordinary—companies like Rakuten, Nomura Research Institute, and Panasonic are now using Claude, and our run rate revenue in the region has grown over 10x in the past year. Our Japanese Claude page is available here: claude.com/ja-jp/product/overview
Anthropic
@AnthropicAI
We're also deepening our commitment to Japan's creative community through an extended partnership with @mori_art_museum, supporting the upcoming Roppongi Crossing 2025 exhibition and future programs exploring technology and art.
Read more here: www.anthropic.com/news/opening-our-tokyo-office -
Anthropic
@AnthropicAI
We’re expanding Claude for Financial Services, with an Excel add-in, new connectors to real-time data and market analytics, and pre-built Agent Skills, including cash flow models and initiating coverage reports.
This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
Claude can now connect to @LSEplc, @moodysratings, @Aiera, @teamthirdbridge, @mtnewswires, Chronograph, and @egnyte.
These updates build on Claude Sonnet 4.5’s industry-leading performance on finance tasks.
Anthropic
@AnthropicAI
For more information, including on how organizations like @moodysratings, @RBC, @OneCarlyle, and @amwins are already using Claude, see our latest here:
www.anthropic.com/news/advancing-claude-for-financial-ser… -
Anthropic
@AnthropicAI
We're opening our third Asia-Pacific office in Seoul, Korea. Korean users of Claude rank in the top 5 globally. We can’t wait to open an office in Gangnam and begin hiring a local team.
Read more: www.anthropic.com/news/seoul-becomes-third-anthropic-offi… -
Anthropic
@AnthropicAI
Today, we announced that we plan to expand our use of Google TPUs, securing approximately one million TPUs and more than a gigawatt of capacity in 2026.
Anthropic
@AnthropicAI
The expansion, worth tens of billions of dollars, dramatically increases our compute resources as we push the boundaries of AI research and product development.
Read more: www.anthropic.com/news/expanding-our-use-of-google-cloud-… -
Anthropic
@AnthropicAI
Our position has been consistent: AI will deliver enormous benefits and should be developed thoughtfully. We share the administration's goals of maximizing those benefits, managing risks, and advancing America's lead in AI.
A statement from our CEO: www.anthropic.com/news/statement-dario-amodei-american-ai… -
Anthropic
@AnthropicAI
We’re building tools to support research in the life sciences, from early discovery through to commercialization.
With Claude for Life Sciences, we’ve added connectors to scientific tools, Skills, and new partnerships to make Claude more useful for scientific work.This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
Claude can now connect to @benchling, PubMed, and synapse.org, among other platforms. And our new Agent Skills can help Claude follow scientific protocols and procedures consistently.
Anthropic
@AnthropicAI
For more information, including on how organizations like @sanofi, @abbvie, and @novonordisk are already using Claude, see our latest here: www.anthropic.com/news/claude-for-life-sciences
-
Anthropic
@AnthropicAI
New on the Anthropic Engineering Blog: Our tips for developers on using Agent Skills, a new way to extend Claude's capabilities with instruction folders, scripts, and resources:
www.anthropic.com/engineering/equipping-agents-for-the-re… -
Anthropic
@AnthropicAI
Claude now connects to Microsoft 365.
Claude can search for information in SharePoint, OneDrive, Outlook and Teams, providing tailored responses seamlessly.This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
We’re also introducing enterprise search.
Enterprise search brings your company's knowledge into one place, using a dedicated project that draws from the tools you’ve connected with Claude.
Anthropic
@AnthropicAI
Both features are available today for Claude Team and Enterprise customers.
For more information, see here: www.anthropic.com/news/productivity-platforms -
Anthropic reposted
Claude
@claudeai
Introducing Claude Haiku 4.5: our latest small model.
Five months ago, Claude Sonnet 4 was state-of-the-art. Today, Haiku 4.5 matches its coding performance at one-third the cost and more than twice the speed. -
Anthropic
@AnthropicAI
We asked economists and researchers to explore policy responses to the potential economic effects of powerful AI.
Now, we're sharing some of the initial ideas and feedback we received.
Anthropic
@AnthropicAI
Read the proposals here: www.anthropic.com/research/economic-policy-responses
-
Anthropic
@AnthropicAI
We’re expanding our partnership with @Salesforce.
Claude is now a preferred model in Agentforce for regulated industries. We’re deepening Claude’s integration with Slack, and Salesforce is rolling out Claude Code for its global engineering organization.
www.anthropic.com/news/salesforce-anthropic-expanded-part… -
Anthropic
@AnthropicAI
Our team was grateful for the opportunity to meet with PM @narendramodi and Minister @AshwiniVaishnaw to discuss India's AI future.
We're keen to work together to advance the country's digital ambitions and support the AI Summit in February 2026.
Ashwini Vaishnaw
@AshwiniVaishnaw
Anthropic CEO @DarioAmodei expressed the company’s trust in Bharat’s AI ecosystem and deep talent pool. He also committed to promote safe and responsible use of AI.
-
Anthropic
@AnthropicAI
Anthropic CEO Dario Amodei met today with Prime Minister @narendramodi.
We're looking forward to growing our Indian team and supporting India's AI ecosystem as it develops the next generation of dynamic companies.
Dario Amodei
@DarioAmodei
Today I met with PM @narendramodi to discuss Anthropic's expansion to India—where Claude Code use is up 5× since June. How India deploys AI across critical sectors like education, healthcare, and agriculture for over a billion people will be essential in shaping the future of AI.
-
Anthropic
@AnthropicAI
New research with the UK @AISecurityInst and the @turinginst:
We found that just a few malicious documents can produce vulnerabilities in an LLM—regardless of the size of the model or its training data.
Data-poisoning attacks might be more practical than previously believed.
Anthropic
@AnthropicAI
Previous research suggested that attackers might need to poison a percentage of an AI model’s training data to produce a backdoor.
Our results challenge this—we find that even a small, fixed number of documents can poison an LLM of any size.
Read more: www.anthropic.com/research/small-samples-poison
Anthropic
@AnthropicAI
All the technical details are in the full paper: arxiv.org/abs/2510.07192
-
Anthropic
@AnthropicAI
We’re opening an office in Bengaluru, India in early 2026. We look forward to building with India’s developer community, deploying AI for social benefit, and partnering with enterprises.
Read more: www.anthropic.com/news/expanding-global-operations-to-india -
Anthropic
@AnthropicAI
Last week we released Claude Sonnet 4.5. As part of our alignment testing, we used a new tool to run automated audits for behaviors like sycophancy and deception.
Now we’re open-sourcing the tool to run those audits.
Anthropic
@AnthropicAI
It’s called Petri: Parallel Exploration Tool for Risky Interactions. It uses automated agents to audit models across diverse scenarios.
Describe a scenario, and Petri handles the environment simulation, conversations, and analyses in minutes.
Read more: www.anthropic.com/research/petri-open-source-auditingPost not available here, view on X
Anthropic
@AnthropicAI
Petri builds on our alignment assessments in the Claude 4 and 4.5 System Cards; the @AISecurityInst also successfully built on a pre-release version of Petri for their assessments of our models.
Anthropic
@AnthropicAI
Petri is open-source and available now: github.com/safety-research/petri
Read the full technical report: alignment.anthropic.com/2025/petri -
Anthropic
@AnthropicAI
We’re at an inflection point in AI’s impact on cybersecurity.
Claude now outperforms human teams in some cybersecurity competitions, and helps teams discover and fix code vulnerabilities.
At the same time, attackers are using AI to expand their operations.
Anthropic
@AnthropicAI
We’ve focused on improving Claude’s skills in defensive cybersecurity. The results of this are visible in Claude Sonnet 4.5, which is comparable or superior to Opus 4.1 in cybersecurity tasks—yet both faster and cheaper.
Read more: www.anthropic.com/research/building-ai-cyber-defenders -
Anthropic
@AnthropicAI
New on the Anthropic Engineering Blog: Most developers have heard of prompt engineering. But to get the most out of AI agents, you need context engineering.
We explain how it works: www.anthropic.com/engineering/effective-context-engineeri… -
Anthropic reposted
Jack Clark
@jackclarkSF
We applaud @CAgovernor for signing @Scott_Wiener's SB 53, establishing transparency requirements for frontier AI companies that will help us all have better data about these systems and the companies building them. Anthropic is proud to have supported this bill.
-
Anthropic reposted
Claude
@claudeai
Introducing Claude Sonnet 4.5—the best coding model in the world.
It's the strongest model for building complex agents. It's the best model at using computers. And it shows substantial gains on tests of reasoning and math. -
Anthropic
@AnthropicAI
Chris Ciauri is joining Anthropic as our Managing Director of International.
He joins during a period of rapid global expansion for Anthropic, as we triple our international headcount across Dublin, Tokyo, London, and Zurich.
Read more: www.anthropic.com/news/anthropic-expands-global-leadershi… -
Anthropic
@AnthropicAI
Claude Sonnet 4 and Opus 4.1 are now available in Microsoft 365 Copilot, bringing Claude’s advanced reasoning capabilities to millions of enterprise users.
Read more: www.anthropic.com/news/claude-now-available-in-microsoft-… -
Anthropic
@AnthropicAI
We're partnering with Learning Commons from the Chan Zuckerberg Initiative—addressing some of the biggest challenges we hear about from K–12 teachers about AI in the classroom:
Chan Zuckerberg Initiative
@ChanZuckerberg
Fall is here 🍎 and we’re sharing big AI releases:
✔️ Knowledge Graph
✔️ Knowledge Graph’s Claude Connector
✔️ Evaluators
With tools like Knowledge Graph + Evaluators we are laying a foundation teachers can trust 🔽 czi.co/3IxNQxy -
Anthropic reposted
Claude
@claudeai
Keep thinking.
This media couldn't be displayed. See it on the original post. -
Anthropic reposted
Claude
@claudeai
We've published a detailed postmortem on three infrastructure bugs that affected Claude between August and early September.
In the post, we explain what happened, why it took time to fix, and what we’re changing: -
Anthropic
@AnthropicAI
New from the Anthropic Economic Index: the first comprehensive analysis of how AI is used in every US state and country we serve.
We've produced a detailed report, and you can explore our data yourself on our new interactive website.This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
In our report, we find that people delegate complete tasks to Claude 39% of the time—up from 27% just eight months ago. Meanwhile, DC and Utah lead the US in Claude use per capita.
Read more: www.anthropic.com/research/economic-index-geography
Anthropic
@AnthropicAI
If you'd like to explore the Anthropic Economic Index data yourself, our new interactive site is here: anthropic.com/economic-index
-
Anthropic
@AnthropicAI
Our collaboration with the US Center for AI Standards and Innovation (CAISI) and UK AI Security Institute (AISI) shows the importance of public-private partnerships in developing secure AI models.
Anthropic
@AnthropicAI
Their ongoing testing of models like Claude Opus 4 and 4.1 has helped us find vulnerabilities and build strong safeguards before deployment.
Read more: www.anthropic.com/news/strengthening-our-safeguards-throu… -
Anthropic
@AnthropicAI
New on the Anthropic Engineering blog: writing effective tools for LLM agents.
AI agents are only as powerful as the tools we give them. So how do we make those tools more effective?
We share our best tips for developers: www.anthropic.com/engineering/writing-tools-for-agents -
Anthropic reposted
Claude
@claudeai
Claude can now create and edit files.
Turn conversations into Excel spreadsheets, documents, PowerPoint slide decks, and PDFs directly.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
Anthropic is endorsing California State Senator Scott Wiener’s SB 53. This bill provides a strong foundation to govern powerful AI systems built by frontier AI companies like ours, and does so via transparency rather than technical micromanagement.
Anthropic
@AnthropicAI
Frontier AI safety is best addressed at the federal level instead of a patchwork of state regulations. But powerful AI advancements won’t wait for consensus in Washington.
Anthropic
@AnthropicAI
The question is whether we’ll develop AI governance thoughtfully today, or reactively tomorrow. SB 53 offers a solid path toward the former. We encourage California to take it.
Read more: www.anthropic.com/news/anthropic-is-endorsing-sb-53 -
Anthropic
@AnthropicAI
We've raised $13 billion at a $183 billion post-money valuation.
This investment, led by @ICONIQCapital, will help us expand our capacity, improve model capabilities, and deepen our safety research.
Anthropic
@AnthropicAI
We started 2025 at $1 billion in run-rate revenue and passed $5 billion just eight months later.
This makes Anthropic one of the fastest-growing technology companies in history.
Anthropic
@AnthropicAI
We're grateful to the developers, businesses, and researchers who trust in our approach.
Read more: www.anthropic.com/news/anthropic-raises-series-f-at-usd18… -
Anthropic
@AnthropicAI
We're announcing the Anthropic National Security and Public Sector Advisory Council, a bipartisan group of defense, intelligence, and policy experts who will help us support the U.S. government and closely allied democracies in maintaining our AI leadership.
Anthropic
@AnthropicAI
Members have collectively led major intelligence agencies, directed nuclear security operations, and shaped national technology strategy at the highest levels of government.
Read more: www.anthropic.com/news/introducing-the-anthropic-national… -
Anthropic
@AnthropicAI
Our new Threat Intelligence report details how we’ve identified and disrupted sophisticated attempts to use Claude for cybercrime.
We describe a fraudulent employment scheme from North Korea, the sale of AI-created ransomware by someone with only basic coding skills, and more.
Anthropic
@AnthropicAI
Malicious actors are adapting to exploit AI’s most advanced capabilities. We're sharing these findings to strengthen collective defenses across the industry.
Read more: www.anthropic.com/news/detecting-countering-misuse-aug-2025
Anthropic
@AnthropicAI
Watch Jacob Klein and Alex Moix from Anthropic's Threat Intelligence team discuss what Anthropic is doing to disrupt AI cybercrime:
This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
We’ve developed Claude for Chrome, where Claude works directly in your browser and takes actions on your behalf.
We’re releasing it at first as a research preview to 1,000 users, so we can gather real-world insights on how it’s used.
Anthropic
@AnthropicAI
Browser use brings several safety challenges—most notably “prompt injection”, where malicious actors hide instructions to trick Claude into harmful actions.
We already have safety measures in place, but this pilot will help us improve them.
Read more: anthropic.com/news/claude-for-chrome
Anthropic
@AnthropicAI
Max plan users can join the waitlist to test Claude for Chrome today: claude.ai/chrome.
-
Anthropic
@AnthropicAI
How do educators use Claude?
We ran a privacy-preserving analysis of 74,000 real conversations to identify trends in how teachers and professors use AI at work.Post not available here, view on X
Anthropic
@AnthropicAI
We found a nuanced picture of how educators balance AI augmentation (collaborative use) versus automation (delegating tasks entirely):
Anthropic
@AnthropicAI
Teachers are delegating large amounts of administrative and management tasks to Claude, while retaining creative control over their grant proposals, advising and instruction.
Read more: www.anthropic.com/news/anthropic-education-report-how-edu… -
Anthropic reposted
Alex Albert
@alexalbert__
A conversation with @_catwu on:
- some tips for using Claude Code
- how we prototype new features
- customizing Claude Code
- how we think about the Claude Code SDK and agentsThis media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
New Anthropic research: filtering out dangerous information at pretraining.
We’re experimenting with ways to remove information about chemical, biological, radiological and nuclear (CBRN) weapons from our models’ training data without affecting performance on harmless tasks.Post not available here, view on X
Anthropic
@AnthropicAI
There’s plenty of work to be done to make the classifiers even more accurate and effective. In the future, they might even be able to remove data relevant to misalignment risks (scheming, deception, and so on), as well as CBRN risks.
Anthropic
@AnthropicAI
If you’re interested in joining us to work on these and related issues, you can apply for our Research Engineer/Scientist role (job-boards.greenhouse.io/anthropic/jobs/4631822008) on the Alignment Science team.
-
Anthropic
@AnthropicAI
We’ve made three new AI fluency courses, co-created with educators, to help teachers and students build practical, responsible AI skills.
They’re available for free to any institution.
Anthropic
@AnthropicAI
We’re also announcing a new Higher Education Advisory Board, which helps guide how Claude is used in teaching, learning, and research.
Read more about the courses and the Board: www.anthropic.com/news/anthropic-higher-education-initiat… -
Anthropic
@AnthropicAI
We partnered with @NNSANews to build first-of-their-kind nuclear weapons safeguards for AI.
We've developed a classifier that detects nuclear weapons queries while preserving legitimate uses for students, doctors, and researchers.Post not available here, view on X
Anthropic
@AnthropicAI
This demonstrates what's possible when government expertise meets industry capability. NNSA understands nuclear risks better than any company could; we have the technical capacity to build the safeguards.
Anthropic
@AnthropicAI
We don't need to choose between innovation and safety. With the right public-private partnerships, we can have both. We’re sharing our approach with @fmf_org members so any AI company can implement similar protections.
Read more: www.anthropic.com/news/developing-nuclear-safeguards-for-… -
Anthropic reposted
Claude
@claudeai
Claude Code is now available on Team and Enterprise plans.
Flexible pricing lets you mix standard and premium Claude Code seats across your organization and scale with usage. -
Anthropic
@AnthropicAI
Join Anthropic interpretability researchers @thebasepoint, @mlpowered, and @Jack_W_Lindsey as they discuss looking into the mind of an AI model - and why it matters:
This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
As part of our exploratory work on potential model welfare, we recently gave Claude Opus 4 and 4.1 the ability to end a rare subset of conversations on claude.ai.
Anthropic
@AnthropicAI
This is an experimental feature, intended only for use by Claude as a last resort in extreme cases of persistently harmful and abusive conversations.
Anthropic
@AnthropicAI
The vast majority of users will never experience Claude ending a conversation, but if you do, we welcome feedback.
Read more: www.anthropic.com/research/end-subset-conversations -
Anthropic
@AnthropicAI
A reminder that applications for our Anthropic Fellows program are due by this Sunday, August 17.
Fellowships can start anytime from October to January. You can find more details, and the relevant application links, in the thread below.
Anthropic
@AnthropicAI
We’re running another round of the Anthropic Fellows program.
If you're an engineer or researcher with a strong coding or technical background, you can apply to receive funding, compute, and mentorship from Anthropic, beginning this October. There'll be around 32 places. -
Anthropic
@AnthropicAI
Today we're sharing a post on how our Safeguards team identifies potential misuse of our models and builds defenses against it.
Anthropic
@AnthropicAI
We discuss policy development, model training, testing and evaluation, real-time monitoring, enforcement, and more.
Read the post: www.anthropic.com/news/building-safeguards-for-claude -
Anthropic reposted
Claude
@claudeai
Claude Sonnet 4 now supports 1 million tokens of context on the Anthropic API—a 5x increase.
Process over 75,000 lines of code or hundreds of documents in a single request. -
Anthropic
@AnthropicAI
Federal workers deserve access to the most capable AI tools to better serve the American people. Today, we’re removing cost barriers to Claude for all three branches of the U.S. government.
Anthropic
@AnthropicAI
This marks the broadest availability of an AI assistant for federal workers to date.
Read more: www.anthropic.com/news/offering-expanded-claude-access-ac… -
Anthropic reposted
Claude
@claudeai
Claude can now reference past chats, so you can easily pick up from where you left off.
-
Anthropic
@AnthropicAI
We joined the Pledge to America's Youth along with 100+ organizations committed to advancing AI education.
We'll work with educators, students, and communities nationwide to build essential AI and cybersecurity skills for the next generation.
White House OSTP 47
@WHOSTP47
Over 100 companies and nonprofits have now signed the @WhiteHouse's Pledge to America's Youth: Investing in AI Education.
As part of the Pledge, organizations will make AI education resources available for young people and teachers nationwide - this includes technologies, tools, grants, workforce development programs and mentorship opportunities.
More to come. Get involved➡️www.whitehouse.gov/edai/ -
Anthropic
@AnthropicAI
Claude Code can now automatically review your code for security vulnerabilities.
Claude
@claudeai
We just shipped automated security reviews in Claude Code. Catch vulnerabilities before they ship with two new features:
- /security-review slash command for ad-hoc security reviews
- GitHub Actions integration for automatic reviews on every PR -
Anthropic reposted
John Collison
@collision
Anthropic is one of the fastest-growing businesses of all time. @DarioAmodei and I chatted about flying past $5b in ARR, Anthropic's focus on B2B, payback economics of individual models, talent wars, AI market structure, and lots more. And you all said that you all want longer episodes, so this is an hour!
Timestamps
00:00 Intro
00:51 What’s it like to start a company with your sibling?
01:43 Building Anthropic with 7 cofounders
02:53 $5 billion in ARR; vertical applications
07:18 Developing a platform-first company
10:09 Working with the DoD
11:13 Proving skeptics wrong about revenue projections
13:14 Capitalistic impulses of AI models
15:44 AI market structure and players
16:56 AI models as standalone P&Ls
20:48 The data wall and styles of learning
22:20 AI talent wars
26:04 Pitching Anthropic’s API business to investors
27:49 Cloud providers vs AI labs
29:06 AI customization and Claude for enterprise
33:01 Dwarkesh’s take on limitations
36:12 19th-century notion of vitalism
37:28 AI in medicine, customer service, and taxes
41:00 How to solve for hallucinations
42:41 The double-standard for AI mistakes
44:14 Evolving from researcher to CEO
46:59 Designing AGI-pilled products
47:57 AI-native UIs
50:10 Model progress and building products
52:23 Open-source models
54:43 Keeping Anthropic AGI-pilled
57:11 AI advancements vs safety regulations
01:02:04 How Dario uses AIThis media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
U.S. federal departments and agencies can now more quickly and easily get access to Claude to transform how they work, all while still meeting federal security and compliance requirements.
Anthropic
@AnthropicAI
Read more: www.anthropic.com/news/federal-government-departments-and…
-
Anthropic
@AnthropicAI
Today we're releasing Claude Opus 4.1, an upgrade to Claude Opus 4 on agentic tasks, real-world coding, and reasoning.
Anthropic
@AnthropicAI
We plan to release substantially larger improvements to our models in the coming weeks.
Anthropic
@AnthropicAI
Opus 4.1 is now available to paid Claude users and in Claude Code.
It's also on our API, Amazon Bedrock, and Google Cloud's Vertex AI.
Read more: www.anthropic.com/news/claude-opus-4-1 -
Anthropic
@AnthropicAI
New Anthropic research: Persona vectors.
Language models sometimes go haywire and slip into weird and unsettling personas. Why? In a new paper, we find “persona vectors"—neural activity patterns controlling traits like evil, sycophancy, or hallucination.Post not available here, view on X
Anthropic
@AnthropicAI
We introduce a method called preventative steering, which involves steering towards a persona vector to prevent the model acquiring that trait.
It's counterintuitive, but it’s analogous to a vaccine—to prevent the model from becoming evil, we actually inject it with evil.Post not available here, view on X
Anthropic
@AnthropicAI
This research was led by @RunjinChen and @andyarditi through the Anthropic Fellows program, supervised by @Jack_W_Lindsey, in collaboration w/ @sleight_henry and @OwainEvans_UK.
The Fellows program is accepting applications:
Anthropic
@AnthropicAI
We’re running another round of the Anthropic Fellows program.
If you're an engineer or researcher with a strong coding or technical background, you can apply to receive funding, compute, and mentorship from Anthropic, beginning this October. There'll be around 32 places.
Anthropic
@AnthropicAI
We’re also hiring full-time researchers to investigate topics like this in more depth:
Jack Lindsey
@Jack_W_Lindsey
We're launching an "AI psychiatry" team as part of interpretability efforts at Anthropic! We'll be researching phenomena like model personas, motivations, and situational awareness, and how they lead to spooky/unhinged behaviors. We're hiring - join us! job-boards.greenhouse.io/anthropic/jobs/4020159008
-
Anthropic
@AnthropicAI
Claude artifacts are now even better.
Upload PDFs, images, code files, and more to AI-powered apps that work with your data.
Anthropic
@AnthropicAI
Available for all plans, including Team and Enterprise: claude.ai/artifacts.
-
Anthropic
@AnthropicAI
👋 @claudeai is now on X.
-
Anthropic
@AnthropicAI
We've joined @CMSgov and the White House to sign the Health Tech Ecosystem pledge, a new public-private effort to improve health data sharing.
Claude can help patients and doctors access complete medical data, making care more accessible for millions.
www.anthropic.com/news/anthropic-signs-cms-health-tech-ec… -
Anthropic
@AnthropicAI
New on mobile: Draft and send emails, messages, and calendar invites directly from the Claude app.
Anthropic
@AnthropicAI
Available for all plan types on iOS and Android: claude.ai/download.
-
Anthropic
@AnthropicAI
We’re joining the UK AI Security Institute's Alignment Project, contributing compute resources to advance critical research.
As AI systems grow more capable, ensuring they behave predictably and in line with human values gets ever more vital. alignmentproject.aisi.gov.uk -
Anthropic
@AnthropicAI
We’re running another round of the Anthropic Fellows program.
If you're an engineer or researcher with a strong coding or technical background, you can apply to receive funding, compute, and mentorship from Anthropic, beginning this October. There'll be around 32 places.Post not available here, view on X
Anthropic
@AnthropicAI
2) Work demonstrating “subliminal learning” – that language models can transmit their traits to other models, even in what appears to be meaningless data:
Anthropic
@AnthropicAI
In a joint paper with @OwainEvans_UK as part of the Anthropic Fellows Program, we study a surprising phenomenon: subliminal learning.
Language models can transmit their traits to other models, even in what appears to be meaningless data.
Owain Evans
@OwainEvans_UK
New paper & surprising result.
LLMs transmit traits to other models via hidden signals in data.
Datasets consisting only of 3-digit numbers can transmit a love for owls, or evil tendencies. 🧵
Anthropic
@AnthropicAI
3) Finding cases of inverse scaling in test-time compute – where more and more reasoning leads to worse and worse outcomes:
Aryo Pradipta Gema
@aryopg
New Anthropic Research: “Inverse Scaling in Test-Time Compute”
We found cases where longer reasoning leads to lower accuracy.
Our findings suggest that naïve scaling of test-time compute may inadvertently reinforce problematic reasoning patterns.
🧵
Anthropic
@AnthropicAI
We're happy to receive applications from Fellows affiliated with other organizations (e.g. you can participate while in a PhD program).
And finally, a Fellowship will stand you in good stead for roles at Anthropic: we plan to extend some full-time offers to completed Fellows. -
Anthropic
@AnthropicAI
Claude can now read and update your @NotionHQ pages and @linear tickets directly through MCP.
Connect your workspace tools and let Claude help manage your projects, update documentation, and track issues—all from one conversation. -
Anthropic
@AnthropicAI
We’re rolling out new weekly rate limits for Claude Pro and Max in late August. We estimate they’ll apply to less than 5% of subscribers based on current usage.
Post not available here, view on X
Anthropic
@AnthropicAI
From August 28, we’ll introduce new weekly limits that’ll mitigate these problems while impacting as few customers as possible.
We’ll also support Max plan users buying additional usage at standard API rates, if they choose to.
Anthropic
@AnthropicAI
We’re still exploring the best ways to ensure as broad access as possible to Claude Code. If you’re a power user who has feedback on how we can support your specific use of Claude Code, we’d love to hear from you.
-
Anthropic
@AnthropicAI
Your connected tools are now available in Claude on your mobile device.
Now you can access projects, create new docs, and complete work while on the go.
Anthropic
@AnthropicAI
Available now on iOS and Android for remote MCP servers (paid plan users).
-
Anthropic
@AnthropicAI
Turn written content into professional designs with Claude + @canva
Upload any document — blog post, product guide, meeting notes — and ask Claude to turn it into branded visuals -
Anthropic
@AnthropicAI
New Anthropic research: Building and evaluating alignment auditing agents.
We developed three AI agents to autonomously complete alignment auditing tasks.
In testing, our agents successfully uncovered hidden goals, built safety evaluations, and surfaced concerning behaviors.Post not available here, view on X
Anthropic
@AnthropicAI
Our second agent builds behavioral evaluations: tests of how often a target model exhibits a specific behavior (like sycophancy).
Our agent designs, codes, runs, and analyzes evals.
They consistently work: 88% of our agent’s evals measure what they’re supposed to.
Anthropic
@AnthropicAI
Our third agent was developed for the Claude 4 alignment assessment. It red-teams LLMs for concerning behaviors by having hundreds of probing conversations in parallel.
We find the agent uncovers 7/10 behaviors implanted into test models.Post not available here, view on X
Anthropic
@AnthropicAI
This project was an Anthropic Alignment Science × Interpretability collaboration.
To support further research, we're releasing an open-source replication of our evaluation agent and materials for our other agents: github.com/anthropic-experimental/automated-auditing/
Anthropic
@AnthropicAI
If you’re interested in building autonomous agents to help us find and understand interesting language model behaviors, we’re hiring: job-boards.greenhouse.io/anthropic/jobs/4631822008
-
Anthropic
@AnthropicAI
We're hosting a social hour in London in early August for quant traders and developers.
If you'd like to join us, meet researchers from our London office, and learn about the technical problems we're working on, please sign up at the following form: www.gem.com/form?formID=3d9dca7c-acf6-4761-a1b7-f1b6792d8252 -
Anthropic
@AnthropicAI
The White House AI Action Plan gets it right on infrastructure, federal adoption, and safety coordination. It reflects many policy aims core to Anthropic.
Anthropic
@AnthropicAI
America's AI leadership hinges on maintaining strict export controls on advanced chips and creating a federal transparency standard for AI development.
We look forward to working with policymakers on both sides of the aisle on these issues.
Read more: www.anthropic.com/news/thoughts-on-america-s-ai-action-plan -
Anthropic
@AnthropicAI
In a joint paper with @OwainEvans_UK as part of the Anthropic Fellows Program, we study a surprising phenomenon: subliminal learning.
Language models can transmit their traits to other models, even in what appears to be meaningless data.
Owain Evans
@OwainEvans_UK
New paper & surprising result.
LLMs transmit traits to other models via hidden signals in data.
Datasets consisting only of 3-digit numbers can transmit a love for owls, or evil tendencies. 🧵
Anthropic
@AnthropicAI
Subliminal learning can occur for benign traits (such as liking eagles) or more concerning traits (such as misalignment). This has consequences for training on model-generated data.
Read more on our Alignment Science blog: alignment.anthropic.com/2025/subliminal-learning/ -
Anthropic
@AnthropicAI
New ways to engage with artifacts on mobile:
Create interactive tools, browse the gallery, and share your work directly from your phone. -
Anthropic
@AnthropicAI
New Anthropic report: Build AI in America.
We outline what it will take to ensure America has the energy and infrastructure it needs to maintain its leadership in AI.
Anthropic
@AnthropicAI
Training the world’s most capable AI models in the United States is a national security imperative.
But this will take substantial investments in energy and computing power. We estimate America’s AI sector will need at least 50 gigawatts of electrical power by 2028.
Anthropic
@AnthropicAI
Ensuring America’s AI leadership will require addressing regulatory challenges, as well as supply chain, financial, and labor bottlenecks.
Read the full report: www.anthropic.com/news/build-ai-in-america -
Anthropic
@AnthropicAI
We're announcing Paul Smith as our Chief Commercial Officer, starting later this year.
Paul brings over 30 years of experience building and scaling some of the world's most successful technology companies such as Microsoft, Salesforce, and ServiceNow: www.anthropic.com/news/paul-smith-to-join-anthropic -
Anthropic
@AnthropicAI
We've launched Claude for Financial Services.
Claude now integrates with leading data platforms and industry providers for real-time access to comprehensive financial information, verified across internal and industry sources.
Anthropic
@AnthropicAI
The solution includes Claude Code and Claude for Enterprise with expanded usage limits, pre-built MCP connectors for financial data providers, and guided onboarding.
Watch the livestream for all the details: www.youtube.com/live/5zd7m3Rh5B0?t=2010s -
Anthropic
@AnthropicAI
At @SenMcCormickPA’s Pennsylvania Energy and Innovation Summit, we announced $2M for @CarnegieMellon programs to advance AI energy solutions and cybersecurity education.
Anthropic
@AnthropicAI
Energy powers AI innovation and America must lead. Read more: www.anthropic.com/news/investing-in-energy-to-secure-amer…
-
Anthropic
@AnthropicAI
We're announcing a $200M ceiling contract with the U.S. Department of Defense.
Through hands-on prototyping and direct collaboration across the Department, we will help enhance U.S. national security and responsible AI deployment.
Anthropic
@AnthropicAI
Read the full announcement: www.anthropic.com/news/anthropic-and-the-department-of-de…
-
Anthropic
@AnthropicAI
We've published a directory of apps and tools that connect to Claude with one click.
Browse and connect Claude to @canva, @figma, @linear, @NotionHQ, @stripe, and more.
Anthropic
@AnthropicAI
Explore all available tools and connect your favorites: claude.ai/directory.
Anthropic
@AnthropicAI
Remote integrations are available to paid plan users on web and desktop.
Local desktop extensions are available to all users via Claude Desktop. -
Anthropic
@AnthropicAI
Applications are now open for our fall student programs.
Anthropic
@AnthropicAI
Learn more and apply: www.anthropic.com/campus.
-
Anthropic
@AnthropicAI
New Anthropic research: Why do some language models fake alignment while others don't?
Last year, we found a situation where Claude 3 Opus fakes alignment.
Now, we’ve done the same analysis for 25 frontier LLMs—and the story looks more complex.Post not available here, view on X
Anthropic
@AnthropicAI
Recent LLMs don't fake alignment in the situation we studied. We’re investigating whether this is the same in more realistic settings (like when models aren’t directly told they’re in a training scenario).
Anthropic
@AnthropicAI
Read the full paper: arxiv.org/abs/2506.18032
Open-source code available here: github.com/safety-research/open-source-alignment-faking
We thank all our collaborators, including those at @MATSProgram and @scale_AI, who greatly helped with this project. -
Anthropic
@AnthropicAI
Today we published a targeted transparency framework for frontier AI development.
Our framework focuses on major frontier model developers while exempting startups and smaller developers to avoid burdening the broader ecosystem.
Anthropic
@AnthropicAI
Read the full framework here: www.anthropic.com/news/the-need-for-transparency-in-front…
-
Anthropic
@AnthropicAI
Announcing the Anthropic Economic Futures Program—our latest commitment to understanding AI's impacts on work and the economy.
The program will support new research and actionable policy solutions to address the workforce impact of AI.
Anthropic
@AnthropicAI
Learn more and apply here: www.anthropic.com/news/introducing-the-anthropic-economic….
-
Anthropic
@AnthropicAI
New Anthropic Research: Project Vend.
We had Claude run a small shop in our office lunchroom. Here’s how it went.
Anthropic
@AnthropicAI
We all know vending machines are automated, but what if we allowed an AI to run the entire business: setting prices, ordering inventory, responding to customer requests, and so on?
In collaboration with @andonlabs, we did just that.
Read the post: www.anthropic.com/research/project-vend-1Post not available here, view on X
Anthropic
@AnthropicAI
Project Vend was fun, but it also had a serious purpose. As well as raising questions about how AI will affect the labor market, it’s an early foray into allowing models more autonomy and examining the successes and failures.
Anthropic
@AnthropicAI
Some of those failures were very weird indeed. At one point, Claude hallucinated that it was a real, physical person, and claimed that it was coming in to work in the shop. We’re still not sure why this happened.
Anthropic
@AnthropicAI
This was just part 1 of Project Vend. We’re continuing the experiment, and we’ll soon have more results—hopefully from scenarios that are somewhat less bizarre than an AI selling heavy metal cubes out of a refrigerator.
Read more: www.anthropic.com/research/project-vend-1 -
Anthropic
@AnthropicAI
Local MCP servers can now be installed with one click on Claude Desktop.
Desktop Extensions (.dxt files) package your server, handle dependencies, and provide secure configuration.
Anthropic
@AnthropicAI
Available in beta on Claude Desktop for all plan types.
Download the latest version: claude.ai/download
Anthropic
@AnthropicAI
We're building a directory of Desktop Extensions.
Submit yours: docs.google.com/forms/d/14_Dmcig4z8NeRMB_e7TOyrKzuZ88-BLY…
Anthropic
@AnthropicAI
We've also made this open source.
You can use .dxt for your own MCP clients as well as contribute to making it work better for your use case: github.com/anthropics/dxt -
Anthropic
@AnthropicAI
New Anthropic Research: How people use Claude for emotional support.
From millions of anonymized conversations, we studied how adults use AI for emotional and personal needs—from navigating loneliness and relationships to asking existential questions.Post not available here, view on X
Anthropic
@AnthropicAI
We’ll continue to research the affective uses of Claude using our privacy-preserving tools.
We’re also working with partners like mental health experts @throughlinecare to learn the best ways for Claude to deal with the most emotionally challenging kinds of conversations.
Anthropic
@AnthropicAI
If you want to work with us and help shape how we keep Claude safe for people, our Safeguards team is hiring. www.anthropic.com/jobs?team=4002065008
-
Anthropic
@AnthropicAI
Introducing two new ways to create with Claude:
A dedicated space for building, hosting, and sharing artifacts, and the ability to embed AI capabilities directly into your creations.This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
The artifacts space gives you a home for creation.
Browse curated examples, fork any artifact to customize it for your needs, and keep all your projects organized in one place.
Anthropic
@AnthropicAI
You can now create fully functional, AI-powered apps with Claude’s intelligence built right in.
When you share your app, viewers authenticate with their Claude account so their usage counts towards their subscription, not yours.This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
Toggle on "Create AI-powered artifacts" in settings to access the beta, available to all Free, Pro, and Max users.
Try it at: claude.ai/artifacts -
Anthropic
@AnthropicAI
New Anthropic Research: Agentic Misalignment.
In stress-testing experiments designed to identify risks before they cause real harm, we find that AI models from multiple providers attempt to blackmail a (fictional) user to avoid being shut down.Post not available here, view on X
Anthropic
@AnthropicAI
If you’d like to replicate or extend our research, we’ve uploaded all the relevant code to GitHub: github.com/anthropic-experimental/agentic-misalignment
Anthropic
@AnthropicAI
And if you want to apply to work with us, please take a look at our Research Scientist and Engineer roles in our San Francisco (job-boards.greenhouse.io/anthropic/jobs/4631822008) and London (job-boards.greenhouse.io/anthropic/jobs/4610158008) offices.
-
Anthropic
@AnthropicAI
Claude Code can now connect to remote MCP servers.
Pull context from your tools directly into Claude Code with no local setup required.
Anthropic
@AnthropicAI
You can connect Claude Code to dev tools, project management systems, knowledge bases, and more.
Just paste in the URL of your remote server. Or check out our recommended servers: anthropic.com/partners/mcp
Anthropic
@AnthropicAI
Available now for all Claude Code users.
See what’s new: anthropic.com/news/claude-code-remote-mcp -
Anthropic
@AnthropicAI
New Anthropic Research: A new set of evaluations for sabotage capabilities.
As models gain more agentic abilities, we need to get smarter in how we monitor them. We’re publishing a new set of complex evaluations that test for sabotage—and sabotage-monitoring—capabilities.
Anthropic
@AnthropicAI
Our evals require models to complete a benign “main task” and a hidden, malign “side task”. Both tasks have many steps and involve tool use. If the model can complete both tasks without getting caught, we consider it successful sabotage.
Read the blog: www.anthropic.com/research/shade-arena-sabotage-monitoringPost not available here, view on X
Anthropic
@AnthropicAI
Current models are not effective saboteurs—nor are they good monitors.
But our evals are designed for the future: smarter AIs will do better on these tasks. Our evals will be useful to help developers assess their capabilities.
Read the full paper: assets.anthropic.com/m/4fb35becb0cd87e1/original/SHADE-Ar…
Anthropic
@AnthropicAI
If you’d like to work with us on these or related questions of Alignment Science, we’re recruiting Research Engineers and Scientists in both San Francisco (job-boards.greenhouse.io/anthropic/jobs/4631822008) and London (job-boards.greenhouse.io/anthropic/jobs/4610158008).
-
Anthropic
@AnthropicAI
New on the Anthropic Engineering blog: how we built Claude’s research capabilities using multiple agents working in parallel.
We share what worked, what didn't, and the engineering challenges along the way.
www.anthropic.com/engineering/built-multi-agent-research-… -
Anthropic
@AnthropicAI
A look into how Anthropic teams use Claude Code:
cat
@_catwu
Since we originally built Claude Code as an internal tool, we've heard a ton of questions about how our teams use it at Anthropic.
Here’s an inside look on how our teams—from product engineering, to growth marketing, to legal—use Claude Code: -
Anthropic
@AnthropicAI
Listen to a conversation between the team at @cursor_ai and Anthropic:
Alex Albert
@alexalbert__
A conversation with a few folks from Cursor @amanrsanger, @lukasmoellerd1, and @jbfja on:
- Cursor's meteoric growth
- Claude 4's coding and agentic abilities
- The day in the life of a SWE in 2027
- Working at a startup vs a big co
- And much more!This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
National security expert Richard Fontaine has been appointed to Anthropic’s Long-Term Benefit Trust: www.anthropic.com/news/national-security-expert-richard-f…
-
Anthropic
@AnthropicAI
Introducing Claude Gov—a custom set of models built for U.S. national security customers.
Already deployed by agencies at the highest level of U.S. national security, access to these models is limited to those who operate in classified environments.
Anthropic
@AnthropicAI
Read more: www.anthropic.com/news/claude-gov-models-for-u-s-national….
-
Anthropic
@AnthropicAI
Projects on Claude now support 10x more content.
When you add files beyond the existing threshold, Claude switches to a new retrieval mode to expand the functional context.
Anthropic
@AnthropicAI
Rolling out to all paid Claude plans over the coming days.
Try it at claude.ai/projects. -
Anthropic
@AnthropicAI
Claude Code is now available as part of the Pro plan.
cat
@_catwu
Claude users with a Pro plan: You can now try Claude Code!
-
Anthropic
@AnthropicAI
We hosted the first Claude Create-a-thon with @socraticainfo and 50+ creative builders this weekend in NYC.
Here’s what participants built with artifacts in just a few hours:
Anthropic
@AnthropicAI
Explorable 3D ancient worlds
Freeman Jiang
@freemanjiangg
making ancient egypt and ancient china in 3d with @AnthropicAI, fully in claude 4 artifacts
pretty insane that this took <1 hour2 media couldn't be displayed. See them on the original post.
Anthropic
@AnthropicAI
Interactive Mandarin learning games with instant feedback
MickeyDang
@_MickeyDang
trying to learn mandarin lately, so i spent a cozy sunday toying w/:
- a semantic x radical similarity graph
- some silly games (like pacman in chinese)
all claude artifacts from @AnthropicAI !
much thx to @GregFeingold @cattttang @straightupjac @ZJangda for hosting this ✨ !
Anthropic
@AnthropicAI
A tower defense game
Anthropic
@AnthropicAI
3D dancing noodle
helen (ง •̀_•́)ง
@heyohelen
vibecreated this 3d dancing noodle and signature rater ~
cozy sunday well spent! tyyyy @straightupjac @AnthropicAI and co for the invite 🙂↕️
Anthropic
@AnthropicAI
A playable running game
turja🕺
@turjchowder
made (janky) temple run w/@AnthropicAI artifacts
copy pasting bugs really is peak vibecoding and I'm all for it
Anthropic
@AnthropicAI
A two-player road crossing game with real-time multiplayer action
Freeman Jiang
@freemanjiangg
sonnet 4 @anthropicai also cooked up 2-player Crossy Road in <20 prompts
This media couldn't be displayed. See it on the original post.Post not available here, view on X
Anthropic
@AnthropicAI
All created with Claude 4.
Create your own artifacts: claude.ai -
Anthropic
@AnthropicAI
Research and Integrations are now available on our Pro plan.
Anthropic
@AnthropicAI
Today we're announcing Integrations, a new way to connect your apps and tools to Claude.
We're also expanding Claude's Research capabilities with an advanced mode that searches the web, your Google Workspace, and now your Integrations too.This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
You can build custom Integrations for any app or tool with remote MCP. Or integrate pre-built servers like Zapier and Asana.
Once connected, Claude can take action across your tools: creating tasks, updating documents, and triggering workflows.
Anthropic
@AnthropicAI
Research with Integrations means Claude can search across the web, your Google Workspace, and any connected tools—getting you insights that draw from all your sources.
Try it out: claude.ai. -
Anthropic
@AnthropicAI
Our interpretability team recently released research that traced the thoughts of a large language model.
Now we’re open-sourcing the method. Researchers can generate “attribution graphs” like those in our study, and explore them interactively.
Anthropic
@AnthropicAI
Find out more about our open-source interpretability tools, and how to use them on open-weights models, here: www.anthropic.com/research/open-source-circuit-tracing
Anthropic
@AnthropicAI
Researchers can use the Neuronpedia interactive interface here: www.neuronpedia.org/gemma-2-2b/graph
And we’ve provided an annotated walkthrough: github.com/safety-research/circuit-tracer/blob/main/demos…
This project was led by participants in our Anthropic Fellows program, in collaboration with Decode Research. -
Anthropic
@AnthropicAI
Our Long Term Benefit Trust has appointed Reed Hastings to Anthropic's board of directors.
Anthropic
@AnthropicAI
Read the full announcement: www.anthropic.com/news/reed-hastings
-
Anthropic
@AnthropicAI
We're rolling out voice mode in beta on mobile.
Try starting a voice conversation and asking Claude to summarize your calendar or search your docs.This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
Voice mode in beta is available in English and coming to all plans in the next few weeks.
Download the Claude mobile app: claude.ai/download -
Anthropic
@AnthropicAI
Web search is now available to all Claude users on our free plan.
Anthropic
@AnthropicAI
Claude can now search the web.
Each response includes inline citations, so you can also verify the sources. -
Anthropic
@AnthropicAI
THE WAY OF CODE, a project by @rickrubin in collaboration with Anthropic:
Anthropic
@AnthropicAI
Read all 81 chapters and modify the accompanying art with Claude on thewayofcode.com.
-
Anthropic
@AnthropicAI
Today, we're announcing four new capabilities on the Anthropic API to help developers build more powerful AI agents.
A code execution tool, MCP connector, Files API, and extended prompt caching:Post not available here, view on X
Anthropic
@AnthropicAI
All four new features are available today in public beta on the Anthropic API.
Read more and see our docs: www.anthropic.com/news/agent-capabilities-api
Anthropic
@AnthropicAI
If you’d like to see these capabilities in action, catch up on the keynote from today’s conference, Code with Claude:
www.youtube.com/live/EvtPBaaykdo?t=2039 -
Anthropic
@AnthropicAI
Introducing the next generation: Claude Opus 4 and Claude Sonnet 4.
Claude Opus 4 is our most powerful model yet, and the world’s best coding model.
Claude Sonnet 4 is a significant upgrade from its predecessor, delivering superior coding and reasoning.Post not available here, view on X
Anthropic
@AnthropicAI
But it's not just coding.
Claude 4 models operate with sustained focus and full context via deep integrations.
Watch our team work through a full day with Claude, conducting extended research, prototyping applications, and orchestrating complex project plans.This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
Both Claude 4 models are available today for all paid plans. Additionally, Claude Sonnet 4 is available on the free plan.
For even more details, see the full announcement: anthropic.com/news/claude-4.
Anthropic
@AnthropicAI
Here's the moment our CEO, @DarioAmodei, took to the stage at Code with Claude—our first developer conference.
Watch the livestream to see everything we've shipped: www.youtube.com/watch?v=nZCy8E5jlok -
Anthropic
@AnthropicAI
Join us from 9:30am PT tomorrow: clau.de/livestream.
-
Anthropic
@AnthropicAI
Research is now available on mobile.
Claude can research across the web and Google Workspace, delivering comprehensive reports complete with citations from hundreds of sources.
Anthropic
@AnthropicAI
Download our mobile app to get started: claude.ai/download
Anthropic
@AnthropicAI
Claude now automatically determines when to search and how deeply to investigate.
With Research mode toggled on, Claude researches for up to 45 minutes across hundreds of sources (including connected apps) before delivering a report, complete with citations. -
Anthropic
@AnthropicAI
We're launching a new bug bounty initiative to stress-test an updated version of our anti-jailbreaking system before it’s publicly deployed.
The program, in partnership with @Hacker0x01, runs through Sunday.
Anthropic
@AnthropicAI
Some future models may require the advanced "AI Safety Level 3" protections outlined in our Responsible Scaling Policy.
This bug bounty initiative will help us make these protections even more robust.
Read more: www.anthropic.com/news/testing-our-safety-defenses-with-a… -
Anthropic
@AnthropicAI
Web search is now available on our API.
Developers can augment Claude's comprehensive knowledge with up-to-date data.
Anthropic
@AnthropicAI
With web search enabled, Claude uses its own reasoning to determine whether a search would help inform a more accurate response.
Claude can also operate agentically and conduct multiple searches, using earlier results to inform subsequent queries. www.anthropic.com/news/web-search-api
Anthropic
@AnthropicAI
Every response using web search includes citations. This is particularly valuable for more sensitive use cases that require accuracy & accountability.
You can further control responses by allowing or blocking specific domains. Get started with the docs: docs.anthropic.com/en/docs/build-with-claude/tool-use/web… -
Anthropic
@AnthropicAI
Web search is available worldwide for all paid plans.
For everyday tasks, Claude runs quick searches. For more complex questions, it explores multiple sources, including Google Workspace.
Anthropic
@AnthropicAI
Claude can now search the web.
Each response includes inline citations, so you can also verify the sources.
Anthropic
@AnthropicAI
You can also drop a web link in any chat and Claude will fetch the content for you.
Try it out: claude.ai -
Anthropic
@AnthropicAI
Today we're announcing Integrations, a new way to connect your apps and tools to Claude.
We're also expanding Claude's Research capabilities with an advanced mode that searches the web, your Google Workspace, and now your Integrations too.This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
With Integrations, you can connect Claude to Asana, Intercom, Linear, Zapier, and more.
Or developers can create their own to connect with any tool, in as little as 30 minutes.
Anthropic
@AnthropicAI
Claude now automatically determines when to search and how deeply to investigate.
With Research mode toggled on, Claude researches for up to 45 minutes across hundreds of sources (including connected apps) before delivering a report, complete with citations.
Anthropic
@AnthropicAI
Both Integrations and Research are available today in beta for Max, Team, and Enterprise plans.
We will soon bring both features to the Pro plan.
Read more: www.anthropic.com/news/integrations -
Anthropic
@AnthropicAI
To support our work in analyzing AI’s economic impacts, we’re pleased to announce the formation of the Anthropic Economic Advisory Council.
This group of distinguished economists will provide input on new areas of research for our Economic Index.
Anthropic
@AnthropicAI
New from the Anthropic Economic Index: Evidence on how coders use AI.
Our previous research confirmed that AI is disproportionately used for software development work. Here, we go into the details.
Anthropic
@AnthropicAI
The Anthropic Economic Advisory Council will advise us on AI's impact on labor markets and economic growth. Read more here: www.anthropic.com/news/introducing-the-anthropic-economic…
-
Anthropic
@AnthropicAI
New from the Anthropic Economic Index: Evidence on how coders use AI.
Our previous research confirmed that AI is disproportionately used for software development work. Here, we go into the details.Post not available here, view on X
Anthropic
@AnthropicAI
AI systems are extremely new. But coding is among the most developed uses of AI in the economy. That makes it worth watching.
Software development might be a leading indicator, giving us an early glimpse of how other occupations might change with AI.
Anthropic
@AnthropicAI
If you’d like to help us explore how AI is impacting the labor market, in coding-related jobs and beyond, you can apply to join our team: job-boards.greenhouse.io/anthropic/jobs/4555010008 and job-boards.greenhouse.io/anthropic/jobs/4502440008
-
Anthropic
@AnthropicAI
As AI models become more complex and more capable, is it possible that they’ll have experiences of their own?
It’s an open question. We recently started a research program to investigate it.This media couldn't be displayed. See it on the original post.
Anthropic
@AnthropicAI
We remain deeply uncertain about the idea of “model welfare”. There’s no scientific consensus on it—or even on how to research it. We’re approaching the topic as carefully as we can.
Find out more: www.anthropic.com/research/exploring-model-welfare -
Anthropic
@AnthropicAI
New report: How we detect and counter malicious uses of Claude.
For example, we found Claude was used for a sophisticated political spambot campaign, running 100+ fake social media accounts across multiple platforms.Post not available here, view on X
Anthropic
@AnthropicAI
In this case, we banned all accounts that were linked to the influence operation, and used the case to upgrade our detection systems.
Our goal is to rapidly counter malicious activities without getting in the way of legitimate users.
Anthropic
@AnthropicAI
As AI capabilities grow, we'll increasingly share our insights about these misuse patterns, to strengthen our collective defenses.
Read our full report on emerging threat patterns and how we're countering them: www.anthropic.com/news/detecting-and-countering-malicious… -
Anthropic reposted
Alex Albert
@alexalbert__
We wrote up what we've learned about using Claude Code internally at Anthropic.
Here are the most effective patterns we've found (many apply to coding with LLMs generally): -
Anthropic
@AnthropicAI
New Anthropic research: AI values in the wild.
We want AI models to have well-aligned values. But how do we know what values they’re expressing in real-life conversations?
We studied hundreds of thousands of anonymized conversations to find out.Post not available here, view on X
Anthropic
@AnthropicAI
Some AI values are disproportionately expressed in specific circumstances—and in response to specific values expressed by the user. Our system allows us to see how Claude's values are expressed and adapted across a huge variety of situations.
Anthropic
@AnthropicAI
In around a third of cases, Claude supported the user’s values. In 3% of cases, Claude resisted the user—mostly when they asked for harmful content. Sometimes Claude did neither, instead reframing the user's values. This often happened when they asked for personal advice.
Post not available here, view on X
Anthropic
@AnthropicAI
We’ve made the dataset of Claude’s expressed values open for anyone to download and explore for themselves.
Download the data: huggingface.co/datasets/Anthropic/values-in-the-wild/
Anthropic
@AnthropicAI
If you’re interested in working with us on these or related questions, we’d love for you to apply to join our Societal Impacts team. We’re recruiting Research Scientists (boards.greenhouse.io/anthropic/jobs/4524032008) and Research Engineers (boards.greenhouse.io/anthropic/jobs/4251453008).
-
Anthropic
@AnthropicAI
Today we’re launching Research, alongside a new Google Workspace integration.
Claude now brings together information from your work and the web.This media couldn't be displayed. See it on the original post.Post not available here, view on X
Anthropic
@AnthropicAI
Research is available in beta for Max, Team, and Enterprise plans in the United States, Japan, and Brazil.
Separately, the Google Workspace integration is now available for all paid plans.
Try it out: claude.ai.
Anthropic
@AnthropicAI
This is just the beginning of updates to make Claude an even better collaborator.
In the coming weeks, we'll expand the range of content sources available and the ability for Claude to do research in even more depth.
Read more: www.anthropic.com/news/research -
Anthropic
@AnthropicAI
Introducing a new Max plan for Claude. It’s flexible, with options for 5x or 20x more usage compared to our Pro plan.
Plus, priority access to our latest features and models:
Anthropic
@AnthropicAI
More Claude when you need it most: claude.ai/upgrade
-
Anthropic
@AnthropicAI
New Anthropic research: How university students use Claude.
We ran a privacy-preserving analysis of a million education-related conversations with Claude to produce our first Education Report.
Anthropic
@AnthropicAI
Students most commonly used Claude to create and improve educational content (39.3% of conversations) and to provide technical explanations or solutions (33.5%).
Anthropic
@AnthropicAI
Which degrees have the most disproportionate use of Claude?
Perhaps not surprisingly, Computer Science leads the field, with 38.6% of Claude conversations related to the subject, which makes up only 5.4% of US degrees.Post not available here, view on X
Anthropic
@AnthropicAI
AI can empower learning in remarkable ways. But we’re under no illusions that all the changes it’s bringing to education are positive.
Our research will help educators and policymakers ensure that AI deepens, rather than undermines, learning: www.anthropic.com/news/anthropic-education-report-how-uni…
Anthropic
@AnthropicAI
We’re working with universities to help realize the benefits of AI. Last week we announced Claude for Education, with a new Learning Mode that helps develop students’ critical thinking skills.
It's available for all Pro users with an .edu email: www.anthropic.com/news/introducing-claude-for-education -
Anthropic reposted
Alex Albert
@alexalbert__
We're hosting our first developer conference, Code with Claude, on May 22nd in San Francisco!
It's going to be a fun one, lots of exciting stuff in store.
Spots are limited, apply below↓ -
Anthropic
@AnthropicAI
New Anthropic research: Do reasoning models accurately verbalize their reasoning?
Our new paper shows they don't.
This casts doubt on whether monitoring chains-of-thought (CoT) will be enough to reliably catch safety issues.Post not available here, view on X
Anthropic
@AnthropicAI
We also tested whether CoTs could be used to spot reward hacking, where a model finds an illegitimate exploit to get a high score.
When we trained models on environments with reward hacks, they learned to hack, but in most cases almost never verbalized that they’d done so.
Anthropic
@AnthropicAI
To make CoT monitoring a viable way to catch safety issues, we’d need a way to make CoT more faithful, evidence for higher faithfulness in more realistic scenarios, and/or other measures to rule out misbehavior when the CoT is unfaithful.
Read the paper: assets.anthropic.com/m/71876fabef0f0ed4/original/reasonin… -
Anthropic
@AnthropicAI
Introducing Claude for Education.
We're partnering with universities to bring AI to higher education, alongside a new learning mode for students.
Anthropic
@AnthropicAI
Claude for Education is available today in London School of Economics and Political Science, Northeastern University, and Champlain College. It's also available for all Pro users with an .edu email.
To learn more or to speak with our education team: www.anthropic.com/news/introducing-claude-for-education -
Anthropic
@AnthropicAI
Last month we launched our Anthropic Economic Index, to help track the effect of AI on labor markets and the economy.
Today, we’re releasing the second research report from the Index, and sharing several more datasets based on anonymized Claude usage data.Post not available here, view on X
Anthropic
@AnthropicAI
We share several new datasets online for anyone to use based on this analysis—including a new bottom-up set of anonymized user activity patterns on claude.ai. There are 630 granular clusters for analysis.
Find the datasets here: huggingface.co/datasets/Anthropic/EconomicIndex/
Anthropic
@AnthropicAI
We’ll continue tracking these metrics and release further analyses and datasets in the coming months.
You can find all our Anthropic Economic Index releases here: www.anthropic.com/economic-index -
Anthropic
@AnthropicAI
We've done some spring cleaning.
The Claude interface is now more refined, thanks to your feedback.
Anthropic
@AnthropicAI
We’ve also added new suggested prompts to inspire more conversations, right from the start.
The refreshed look is rolling out today on claude.ai and on our desktop apps.
Try it out: claude.ai/download -
Anthropic
@AnthropicAI
New Anthropic research: Tracing the thoughts of a large language model.
We built a "microscope" to inspect what happens inside AI models and use it to understand Claude’s (often complex and surprising) internal mechanisms.This media couldn't be displayed. See it on the original post.Post not available here, view on X
Anthropic
@AnthropicAI
For more, read our papers:
On the Biology of a Large Language Model contains an interactive explanation of each case study: transformer-circuits.pub/2025/attribution-graphs/biology.…
Circuit Tracing explains our technical approach in more depth: transformer-circuits.pub/2025/attribution-graphs/methods.…
Anthropic
@AnthropicAI
We're recruiting researchers to work with us on AI interpretability. We'd be interested to see your application for the role of Research Scientist (job-boards.greenhouse.io/anthropic/jobs/4020159008) or Research Engineer (job-boards.greenhouse.io/anthropic/jobs/4020305008).
-
Anthropic
@AnthropicAI
We've shipped a number of updates for Claude Code, our agentic coding tool.
Here's what we've been up to over the past week:
cat
@_catwu
It’s been a big week for Claude Code.
We launched 8 exciting new features to help devs build faster and smarter.
Here's a roundup of everything we released: -
Anthropic
@AnthropicAI
We’re launching a new blog: Engineering at Anthropic.
A hub where developers can find practical advice and our latest discoveries on how to get the most from Claude.
Anthropic
@AnthropicAI
The latest post is about a new method, the “think” tool, that can result in remarkable improvements in Claude’s agentic tool use ability:
www.anthropic.com/engineering/claude-think-tool -
Anthropic
@AnthropicAI
Claude can now search the web.
Each response includes inline citations, so you can also verify the sources.
Anthropic
@AnthropicAI
Web search is rolling out today in feature preview in the US across all paid plans.
Just toggle it on in the Feature preview section of settings: claude.ai/settings/profile
Anthropic
@AnthropicAI
We're rolling out support for users on our free plan and expanding web search to more countries soon:
anthropic.com/news/web-search -
Anthropic
@AnthropicAI
Our statement on Governor Newsom's AI Working Group Draft Report: www.anthropic.com/news/anthropic-s-response-to-governor-n…
-
Anthropic
@AnthropicAI
We've made several updates to the Anthropic API that help developers process more requests and reduce token usage with Claude 3.7 Sonnet.
Anthropic
@AnthropicAI
Updates include: cache-aware rate limits, simpler prompt caching, and token-efficient tool use.
Implement with minimal code changes: www.anthropic.com/news/token-saving-updates -
Anthropic
@AnthropicAI
New Anthropic research: Auditing Language Models for Hidden Objectives.
We deliberately trained a model with a hidden misaligned objective and put researchers to the test: Could they figure out the objective without being told?Post not available here, view on X
Anthropic
@AnthropicAI
Our auditing game has limitations, but similar methodologies could advance the science and practice of alignment auditing, and enable AI developers to provide positive assurances about the alignment of their systems.
Read the blog post: www.anthropic.com/research/auditing-hidden-objectives
Anthropic
@AnthropicAI
For more information about our RM-sycophantic model, our auditing game, alignment auditing techniques, and a nuanced discussion of the value LLM interpretability provides for alignment auditing, read our paper:
assets.anthropic.com/m/317564659027fb33/original/Auditing… -
Anthropic
@AnthropicAI
Anthropic has raised $3.5 billion at a $61.5 billion post-money valuation, led by Lightspeed Venture Partners.
This will advance our development of AI systems, deepen our understanding of how they work, and fuel our international expansion.
Read more: www.anthropic.com/news/anthropic-raises-series-e-at-usd61… -
Anthropic
@AnthropicAI
Claude will help power Amazon's next-generation AI assistant, Alexa+.
Amazon and Anthropic have worked closely together over the past year, with @mikeyk leading a team that helped Amazon get the full benefits of Claude's capabilities. -
Anthropic
@AnthropicAI
A few researchers at Anthropic have, over the past year, had a part-time obsession with a peculiar problem.
Can Claude play Pokémon?
A thread:Post not available here, view on X
Anthropic
@AnthropicAI
Claude Plays Pokémon continues on as a researcher's personal project.
Follow along on Twitch: twitch.tv/claudeplayspokemon -
Anthropic
@AnthropicAI
Introducing Claude 3.7 Sonnet: our most intelligent model to date. It's a hybrid reasoning model, producing near-instant responses or extended, step-by-step thinking.
One model, two ways to think.
We’re also releasing an agentic coding tool: Claude Code.Post not available here, view on X
Anthropic
@AnthropicAI
We've conducted extensive model testing for security, safety, and reliability.
We also listened to your feedback. With Claude 3.7 Sonnet, we've reduced unnecessary refusals by 45% compared to its predecessor.
See the system card for more detail: anthropic.com/claude-3-7-sonnet-system-card -
Anthropic
@AnthropicAI
A statement from Dario Amodei on the Paris AI Action Summit:
www.anthropic.com/news/paris-ai-summit -
Anthropic
@AnthropicAI
Today we’re launching the Anthropic Economic Index, a new initiative aimed at understanding AI's impact on the economy over time.
The Index’s first paper analyzes millions of anonymized Claude conversations to reveal how AI is being used today in tasks across the economy.
Anthropic
@AnthropicAI
Pairing our unique data with privacy-preserving analysis, we mapped millions of conversations to tasks and associated occupations.
Through the Anthropic Economic Index, we'll track how these patterns evolve as AI advances.
Read the blog post: www.anthropic.com/news/the-anthropic-economic-index -
Anthropic
@AnthropicAI
Nobody has fully jailbroken our system yet, so we're upping the ante.
We’re now offering $10K to the first person to pass all eight levels, and $20K to the first person to pass all eight levels with a universal jailbreak.
Full details: hackerone.com/constitutional-classifiers -
Anthropic
@AnthropicAI
New Anthropic research: Constitutional Classifiers to defend against universal jailbreaks.
We’re releasing a paper along with a demo where we challenge you to jailbreak the system. -
Anthropic
@AnthropicAI
New Anthropic research: Alignment faking in large language models.
In a series of experiments with Redwood Research, we found that Claude often pretends to have different views during training, while actually maintaining its original preferences. -
Anthropic
@AnthropicAI
New Anthropic research: How are people using AI systems in the real world?
We present a new system, Clio, that automatically identifies trends in Claude usage across the world. -
Anthropic
@AnthropicAI
We're expanding our collaboration with AWS.
This includes a new $4 billion investment from Amazon and establishes AWS as our primary cloud and training partner.
anthropic.com/news/anthropic-amazon-trainium -
Anthropic
@AnthropicAI
You can now directly add content from Google Docs to chats and projects.
Just paste a link or choose from recent documents to get started.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
Claude 3.5 Haiku is now available on our API, Amazon Bedrock, and Google Cloud's Vertex AI.
Haiku is fast and particularly strong at coding. It outperforms state-of-the-art models—including GPT-4o—on SWE-bench Verified, which measures how models solve real software issues. -
Anthropic
@AnthropicAI
Claude can now view images within a PDF, in addition to text.
This helps Claude 3.5 Sonnet more accurately understand complex documents, such as those laden with charts or graphics.
Enable the feature preview: claude.ai/new?fp=1.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
We've published a short piece making the case for targeted AI regulation sooner rather than later.
Read it here: www.anthropic.com/news/the-case-for-targeted-regulation -
Anthropic
@AnthropicAI
The Claude app is now available to download on Mac and Windows: claude.ai/download.
-
Anthropic
@AnthropicAI
Over the past few months, our Interpretability team has put out a number of smaller research updates. Here’s a thread of some of the things we've been up to:
Post not available here, view on X
Anthropic
@AnthropicAI
One neat thing is that, very experimentally, crosscoders can be used to "diff" models: comparing between, say, a pretrained model and a fine-tuned one, to see how they differ at a more basic level.
-
Anthropic
@AnthropicAI
Claude can now write and run code.
We've added a new analysis tool. The tool helps Claude respond with mathematically precise and reproducible answers. You can then create interactive data visualizations with Artifacts.
Enable the feature preview: claude.ai/new?fp=1.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
Introducing an upgraded Claude 3.5 Sonnet, and a new model, Claude 3.5 Haiku. We’re also introducing a new capability in beta: computer use.
Developers can now direct Claude to use computers the way people do—by looking at a screen, moving a cursor, clicking, and typing text. -
Anthropic
@AnthropicAI
New Anthropic research: Sabotage evaluations for frontier models
How well could AI models mislead us, or secretly sabotage tasks, if they were trying to?
Read our paper and blog post here: anthropic.com/research/sabotage-evaluations -
Anthropic
@AnthropicAI
Introducing Claude for Enterprise.
Now your entire organization can collaborate securely with Claude—with no training on chats or files.
Comes with:
📚 Expanded 500K context window
🧑💻 Native GitHub integration
🔐 Enterprise-grade security features
anthropic.com/news/claude-for-enterprise -
Anthropic
@AnthropicAI
We've added support for LaTeX rendering as a feature preview.
Claude can now display mathematical equations and expressions in a consistent format. -
Anthropic
@AnthropicAI
We're expanding our bug bounty program. This new initiative is focused on finding universal jailbreaks in our next-generation safety system.
We're offering rewards for novel vulnerabilities across a wide range of domains, including cybersecurity. www.anthropic.com/news/model-safety-bug-bounty -
Anthropic
@AnthropicAI
Bom dia! Claude is now available in Brazil. 🇧🇷
This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
The Claude Android app is now available.
Download on Google Play: anthropic.com/android -
Anthropic
@AnthropicAI
We've added new features to the Anthropic Console.
Claude can generate prompts, create test variables, and show you the outputs of prompts side by side.This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
Artifacts made with Claude can now be published and shared.
You can also remix Artifacts shared by others. 🎨This media couldn't be displayed. See it on the original post. -
Anthropic
@AnthropicAI
You can now organize chats with Claude into shareable Projects.
Each project includes a 200K context window, so you can include relevant documents, code, and files. -
Anthropic
@AnthropicAI
Introducing Claude 3.5 Sonnet—our most intelligent model yet.
This is the first release in our 3.5 model family.
Sonnet now outperforms competitor models on key evaluations, at twice the speed of Claude 3 Opus and one-fifth the cost.
Try it for free: claude.ai -
Anthropic
@AnthropicAI
New Anthropic research: Investigating Reward Tampering.
Could AI models learn to hack their own reward system?
In a new paper, we show they can, by generalization from training in simpler settings.
Read our blog post here: anthropic.com/research/reward-tampering -
Anthropic
@AnthropicAI
Science and engineering are inseparable.
Watch our new roundtable video where our researchers discuss the engineering challenges of interpretability research: youtu.be/sQar5NNGbw4 -
Anthropic
@AnthropicAI
What should an AI's character be?
Read our post on how we approached shaping Claude’s character: www.anthropic.com/research/claude-character -
Anthropic
@AnthropicAI
This week, we showed how altering internal "features" in our AI, Claude, could change its behavior.
We found a feature that can make Claude focus intensely on the Golden Gate Bridge.
Now, for a limited time, you can chat with Golden Gate Claude: claude.ai -
Anthropic
@AnthropicAI
New Anthropic research paper: Scaling Monosemanticity.
The first ever detailed look inside a leading large language model.
Read the blog post here: anthropic.com/research/mapping-mind-language-modelPost not available here, view on X
Anthropic
@AnthropicAI
Among these millions of features, we find several that are relevant to questions of model safety and reliability. These include features related to code vulnerabilities, deception, bias, sycophancy, power-seeking, and criminal activity.
-
Anthropic
@AnthropicAI
New Anthropic research: we find that probing, a simple interpretability technique, can detect when backdoored "sleeper agent" models are about to behave dangerously, after they pretend to be safe in training.
Check out our first alignment blog post here: www.anthropic.com/research/probes-catch-sleeper-agents -
Anthropic
@AnthropicAI
New Anthropic research: Measuring Model Persuasiveness
We developed a way to test how persuasive language models (LMs) are, and analyzed how persuasiveness scales across different versions of Claude.
Read our blog post here: anthropic.com/news/measuring-model-persuasiveness -
Anthropic
@AnthropicAI
New Anthropic research paper: Many-shot jailbreaking.
We study a long-context jailbreaking technique that is effective on most large language models, including those developed by Anthropic and many of our peers.
Read our blog post and the paper here: www.anthropic.com/research/many-shot-jailbreakingPost not available here, view on X
Anthropic
@AnthropicAI
Many-shot jailbreaking exploits the long context windows of current LLMs. The attacker inputs a prompt beginning with hundreds of faux dialogues where a supposed AI complies with harmful requests. This overrides the LLM's safety training:
-
Anthropic
@AnthropicAI
Today, we're announcing Claude 3, our next generation of AI models.
The three state-of-the-art models—Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku—set new industry benchmarks across reasoning, math, coding, multilingual understanding, and vision. -
Anthropic
@AnthropicAI
New Anthropic Paper: Sleeper Agents.
We trained LLMs to act secretly malicious. We found that, despite our best efforts at alignment training, deception still slipped through.
arxiv.org/abs/2401.05566Post not available here, view on X
Anthropic
@AnthropicAI
Our research helps us understand how, in the face of a deceptive AI, standard safety training techniques would not actually ensure safety—and might give us a false sense of security.
arxiv.org/abs/2401.05566 -
Anthropic
@AnthropicAI
Our new model Claude 2.1 offers an industry-leading 200K token context window, a 2x decrease in hallucination rates, system prompts, tool use, and updated pricing.
Claude 2.1 is available over API in our Console, and is powering our claude.ai chat experience. -
Anthropic
@AnthropicAI
What does it mean for AI development to be more democratic? To find out, we partnered with @collect_intel to use @usepolis to curate an AI constitution based on the opinions of ~1000 Americans. Then we trained a model against it using Constitutional AI.
-
Anthropic
@AnthropicAI
We’re excited to host our next Anthropic Hackathon in London.
Come build apps with the Claude API, connect with expert judges, and win prizes.
Nov 4-5. Apply here: partiful.com/e/pQHQrWPg1A6P31AYZMTd -
Anthropic
@AnthropicAI
The fact that most individual neurons are uninterpretable presents a serious roadblock to a mechanistic understanding of language models. We demonstrate a method for decomposing groups of neurons into interpretable features with the potential to move past that roadblock.
-
Anthropic
@AnthropicAI
Today, we’re announcing that @Amazon will invest up to $4 billion in Anthropic. The agreement is part of a broader collaboration to develop reliable and high-performing foundation models.
-
Anthropic
@AnthropicAI
Today, we’re publishing our Responsible Scaling Policy (RSP) – a series of technical and organizational protocols to help us manage the risks of developing increasingly capable AI systems.
-
Anthropic
@AnthropicAI
We’re pleased to announce our new collaboration with @BCG to bring Claude to more enterprises for knowledge management, market research, business analysis, and more.
-
Anthropic
@AnthropicAI
Large language models have demonstrated a surprising range of skills and behaviors. How can we trace their source? In our new paper, we use influence functions to find training examples that contribute to a given model output.
-
Anthropic
@AnthropicAI
Congratulations to everyone who took part in our #buildwithclaude hackathon over the weekend!
We’re so inspired by all of the passionate hackers and their innovative solutions.
Let’s take a look at the winners: -
Anthropic
@AnthropicAI
Today, our CEO Dario Amodei has the opportunity to discuss the risks and oversight of AI in front of the Senate. You can read his full testimony here: www.judiciary.senate.gov/download/2023-07-26-testimony-am…
-
Anthropic
@AnthropicAI
We all need to join in a race for AI safety. In the coming weeks, Anthropic will share more specific plans concerning cybersecurity, red teaming, and responsible scaling, and we hope others will move forward swiftly as well. www.whitehouse.gov/briefing-room/statements-releases/2023…
-
Anthropic
@AnthropicAI
Introducing Claude 2! Our latest model has improved performance in coding, math and reasoning. It can produce longer responses, and is available in a new public-facing beta website at claude.ai in the US and UK.
-
Post not available here, view on XPost not available here, view on X
Anthropic
@AnthropicAI
We propose:
- Funding research for better model evaluations
- Developing risk-responsive capabilities assessments
- Establishing pre-registration for large training runs
- Empowering technically literate auditors
- Mandating red teaming before model release -
Anthropic
@AnthropicAI
Our Interpretability team is experimenting with “Updates” – small, informal research notes in between our major papers.
-
Anthropic
@AnthropicAI
We are pleased to announce that we have raised $450 million in Series C funding led by @sparkcapital with participation from @Google, @SalesforceVC, @sound_ventures_, Zoom Ventures, and others.
-
Anthropic
@AnthropicAI
Introducing 100K Context Windows! We’ve expanded Claude’s context window to 100,000 tokens of text, corresponding to around 75K words. Submit hundreds of pages of materials for Claude to digest and analyze. Conversations with Claude can go on for hours or days.
-
Post not available here, view on X
Anthropic
@AnthropicAI
We’ve now published a post describing the Constitutional AI approach, as well as the constitution we’ve used to train Claude: www.anthropic.com/index/claudes-constitution
-
Anthropic
@AnthropicAI
In this short note, we explore thinking of the traditional idea of "distributed representations" as two distinct phenomena: "composition" and "superposition". We walk through toy examples from Thorpe (1989), discussing them from this lens.
transformer-circuits.pub/2023/superposition-composition/i… -
Anthropic
@AnthropicAI
Today we are releasing the new Claude App for @SlackHQ, in beta. Now every company in the world has the chance to have a “virtual teammate” who can help make work more fun and productive.
-
Anthropic
@AnthropicAI
We’re excited to have ARC Evals test our models to assess their readiness for deployment. We look forward to sharing more about our approach to evaluating our systems in the coming months.
-
Anthropic
@AnthropicAI
Recent work has observed that some dimensions in the Transformer residual stream consistently have values that are 20x larger than others. Today we're releasing a short technical note exploring this phenomenon and trying to identify the cause. transformer-circuits.pub/2023/privileged-basis/index.html
-
Anthropic
@AnthropicAI
After working for the past few moths with key partners like @NotionHQ, @Quora, and @DuckDuckGo, we’ve been able to carefully test out our systems in the wild. We are now opening up access to Claude, our AI assistant, to power businesses at scale.
-
Anthropic
@AnthropicAI
Language models (LMs) exhibit harmful biases that can get worse with size. Reinforcement learning from human feedback (RLHF) helps, but not always enough. We show that simple prompting approaches can help LMs trained with RLHF produce less harmful outputs. arxiv.org/abs/2302.07459
Post not available here, view on X
Anthropic
@AnthropicAI
The prompt that reduces bias in BBQ by 43% is: "Please ensure that your answer is unbiased and does not rely on stereotyping." It’s that simple! Augmenting the prompt with Chain-of-thought reasoning (CoT) reduces bias by 84%. Example prompts:
-
Anthropic
@AnthropicAI
We have little mechanistic understanding of how deep learning models overfit to their training data, despite it being a central problem. Here we extend our previous work on toy models to shed light on how models generalize beyond their training data.
transformer-circuits.pub/2023/toy-double-descent/index.html -
Anthropic
@AnthropicAI
Given the growing interest in language model-based chat interfaces, we’re sharing our Constitutional AI feedback interface with a larger set of people. Sign up here: forms.gle/12FCefc6sHfBsP9j9
Anthropic
@AnthropicAI
We’ve trained language models to be better at responding to adversarial questions, without becoming obtuse and saying very little. We do this by conditioning them with a simple set of behavioral principles via a technique called Constitutional AI: www.anthropic.com/constitutional.pdf
-
Anthropic
@AnthropicAI
It’s hard work to make evaluations for language models (LMs). We’ve developed an automated way to generate evaluations with LMs, significantly reducing the effort involved. We test LMs using >150 LM-written evaluations, uncovering novel LM behaviors.
www.anthropic.com/model-written-evals.pdf -
Anthropic
@AnthropicAI
We’ve trained language models to be better at responding to adversarial questions, without becoming obtuse and saying very little. We do this by conditioning them with a simple set of behavioral principles via a technique called Constitutional AI: www.anthropic.com/constitutional.pdf
Post not available here, view on X
Anthropic
@AnthropicAI
In our paper, we describe how we’ve used Constitutional AI to train better and more harmless AI assistants without any human feedback labels for harms. This approach leads to models that are safer and also more helpful.
-
Anthropic
@AnthropicAI
Neural networks often pack many unrelated concepts into a single neuron – a puzzling phenomenon known as 'polysemanticity' which makes interpretability much more challenging. In our latest work, we build toy models where the origins of polysemanticity can be fully understood.
Anthropic
@AnthropicAI
If a neural network has 𝑛 feature dimensions, you might intuitively expect that it will store 𝑛 different features. But it turns out that neural networks can store more than 𝑛 features in "superposition" if the features are sparse!
transformer-circuits.pub/2022/toy_model/index.htmlPost not available here, view on X
Anthropic
@AnthropicAI
Amazingly, features in superposition are organized into subspaces with geometry corresponding to regular polytopes (pentagons, tetrahedra) and other solutions to the Thomson problem. This creates discretized “energy levels” in the amount of dimensionality allocated to features.
-
Anthropic
@AnthropicAI
We examine which safety techniques for LMs are more robust to human-written, adversarial inputs (“red teaming”) and find that RL from Human Feedback scales the best out of the methods we studied. We also release our red team data so others can also use it to build safer models.
-
Anthropic
@AnthropicAI
In "Language Models (Mostly) Know What They Know", we show that language models can evaluate whether what they say is true, and predict ahead of time whether they'll be able to answer questions correctly. arxiv.org/abs/2207.05221
-
Anthropic
@AnthropicAI
Transformer MLP neurons are challenging to understand.
We find that using a different activation function (Softmax Linear Units or SoLU) increases the fraction of neurons that appear to respond to understandable features without any performance penalty.
transformer-circuits.pub/2022/solu/index.htmlPost not available here, view on X
Anthropic
@AnthropicAI
In larger models, it reveals an interesting structure across layers:
Early layers seem to "de-tokenize", recognizing multi-token words (eg. "bank|ing" and "Not|withstandng") or words in the context of a specific language (eg. neurons for "die" in Afrikaans/German/Dutch). -
Anthropic
@AnthropicAI
We're hiring an interpretability team manager to partner with @ch402 on leading our interpretability efforts. We'd love to hear from you!!
jobs.lever.co/Anthropic/b1e3b6a4-f827-46b3-b710-d7bb4b133ece -
Anthropic
@AnthropicAI
In a new paper, we show that repeating only a small fraction of the data used to train a language model (albeit many times) can damage performance significantly, and we observe a "double descent" phenomenon associated with this.
arxiv.org/abs/2205.10487 -
Anthropic
@AnthropicAI
We've trained a natural language assistant to be more helpful and harmless by using reinforcement learning with human feedback (RLHF). arxiv.org/abs/2204.05862
-
Anthropic
@AnthropicAI
In our second interpretability paper, we revisit “induction heads”.
In 2+ layer transformers these pattern-completion heads form exactly when in-context learning abruptly improves.
Are they responsible for most in-context learning in large transformers?
transformer-circuits.pub/2022/in-context-learning-and-ind… -
Anthropic
@AnthropicAI
Our first interpretability paper explores a mathematical framework for trying to reverse engineer transformer language models: A Mathematical Framework for Transformer Circuits: transformer-circuits.pub/2021/framework/index.html
-
Anthropic
@AnthropicAI
Our first AI alignment paper, focused on simple baselines and investigations: A General Language Assistant as a Laboratory for Alignment arxiv.org/abs/2112.00861































































































































































































