The Warner / Suno settlement is a very good day for musicians, and for people everywhere fighting against exploitative AI.
More details on the agreement are needed, but the bottom line is this: Suno is shifting to models trained on music they've licensed, and shutting down their old, unlicensed ones. This is a huge win.
Does it suck that a company that trained on "essentially all music files of reasonable quality that are accessible on the open internet" - without payment or permission - is rewarded with a major label deal? Of course it does.
But you've got to remember - this was a fight about training data. Suno said they could train on musicians' work without permission - the record labels said they had to license it. And Suno is now going to license it. That alone is a huge win.
And remember - it was only thanks to this lawsuit that Suno admitted what they trained on in the first place. Before the lawsuit, they were asked several times about their training data, and prevaricated every time. It was only thanks to this lawsuit that they ended up admitting that their models were trained on musicians' work without permission.
So the lawsuit has led to them admitting what they trained on, *and* licensing the music in question. That is a great outcome, and shows the importance of AI copyright lawsuits in the fight against the exploitation of creatives by AI companies.
Yes, we should be wary of celebrating too much when, of course, Suno trained on lots of music that Warner doesn't own, and that is therefore not covered by this settlement. But they have not settled with the other major labels yet, and they are being sued by others, too - for instance, in two class action lawsuits brought by independent musicians. This settlement doesn't get them off the hook.
What it *does* do is continue the gradual shift away from a wild west approach in which AI companies use whatever creative content they want, irrespective of copyright, and towards a world in which AI companies pay for their training data.
So when AI boosters tell you that it's too late to do anything about the exploitation at the heart of generative AI - when they tell you Pandora's box is open - point them to this settlement.
When they tell you AI companies will never have to get opt-in consent - point them to this settlement, which says otherwise.
There is still lots of work to be done in the cause of trying to get fairness for creatives in the age of AI. Licensing training data isn't everything - we also need to protect creatives' work, label AI outputs, avoid slop polluting the internet, and much more besides.
But this is a great step. We are winning. The fight is worth it. Keep fighting.