There are dozens of comparable copyright lawsuits working through the courts without delay, with cases filed against all the highest players—not only Anthropic and Meta but Google, OpenAI, Microsoft, and more. On the opposite side, plaintiffs range from individual artists and authors to large firms like Getty and the Latest York Times.
The outcomes of those cases are set to have an infinite impact on the longer term of AI. In effect, they may determine whether or not model makers can proceed ordering up a free lunch. If not, they may need to begin paying for such training data via recent sorts of licensing deals—or find recent ways to coach their models. Those prospects could upend the industry.
And that’s why last week’s wins for the technology firms matter. So: Cases closed? Not quite. In case you drill into the main points, the rulings are less cut-and-dried than they appear at first. Let’s take a better look.
In each cases, a bunch of authors (the Anthropic suit was a category motion; 13 plaintiffs sued Meta, including high-profile names equivalent to Sarah Silverman and Ta-Nehisi Coates) got down to prove that a technology company had violated their copyright through the use of their books to coach large language models. And in each cases, the businesses argued that this training process counted as fair use, a legal provision that allows using copyrighted works for certain purposes.
There the similarities end. Ruling in Anthropic’s favor, senior district judge William Alsup argued on June 23 that the firm’s use of the books was legal because what it did with them was transformative, meaning that it didn’t replace the unique works but made something recent from them. “The technology at issue was amongst essentially the most transformative lots of us will see in our lifetimes,” Alsup wrote in his judgment.
In Meta’s case, district judge Vince Chhabria made a special argument. He also sided with the technology company, but he focused his ruling as an alternative on the problem of whether or not Meta had harmed the marketplace for the authors’ work. Chhabria said that he thought Alsup had brushed aside the importance of market harm. “The important thing query in virtually any case where a defendant has copied someone’s original work without permission is whether or not allowing people to have interaction in that kind of conduct would substantially diminish the marketplace for the unique,” he wrote on June 25.
Same end result; two very different rulings. And it’s not clear exactly what meaning for the opposite cases. On the one hand, it bolsters not less than two versions of the fair-use argument. On the opposite, there’s some disagreement over how fair use ought to be decided.
But there are even greater things to notice. Chhabria was very clear in his judgment that Meta won not since it was in the best, but since the plaintiffs didn’t make a powerful enough argument. “Within the grand scheme of things, the implications of this ruling are limited,” he wrote. “This shouldn’t be a category motion, so the ruling only affects the rights of those 13 authors—not the countless others whose works Meta used to coach its models. And, as should now be clear, this ruling doesn’t stand for the proposition that Meta’s use of copyrighted materials to coach its language models is lawful.” That reads so much like an invite for anyone else on the market with a grievance to come back and have one other go.
