AI & Copyright: Federal Judges Address “Fair Use” Issue

Last week, California federal district court judges issued two opinions addressing the critical issue of whether the use of copyrighted materials to train AI models constituted a “fair use” of those materials.  This intro to Sullivan & Cromwell’s memo on the decisions summarizes the judges’ rulings:

On June 23, 2025, the U.S. District Court for the Northern District of California (Judge Alsup) held in Bartz v. Anthropic that Anthropic’s use of copyrighted books to train its generative artificial intelligence (“AI”) models was protected by fair use under U.S. copyright law. In particular, Judge Alsup emphasized that Anthropic’s use was “spectacularly” and “exceedingly” transformative. Two days later, on June 25, a different judge of the same court (Judge Chhabria) held in Kadrey v. Meta Platforms that Meta’s use of copyrighted books to train its models was likewise protected by fair use, finding that there was “no serious question” that Meta’s use of copyrighted materials was “highly transformative.”

These are the first decisions by U.S. federal courts to address whether a central feature of many AI companies’ business models—the use of copyrighted materials to train generative AI models—infringes the copyrights in such materials. Whether these opinions survive on appeal or what parts, if any, of their reasoning are followed by other courts facing similar questions remains to be seen. These early cases, however, represent a positive development for the AI industry seeking clarity on these significant questions.

However, the judge in the Bartz case went on to hold that the also held that Anthropic’s use of unauthorized copies of copyrighted materials to build a central library was not fair use.  This excerpt from Judge Alsup’s order summarizes his position with respect to the use of “pirated” material:

The copies used to build a central library and that were obtained from pirated sources
plainly displaced demand for Authors’ books — copy for copy. Not every person who merely
intends to make a fair use of a work is thereby entitled to a full copy in the meantime, nor even to steal a copy so that achieving this fair use is especially simple or cost-effective. Here, the copies employed in training LLMs were one thing, but the copies acquired to assemble a convenient, general-purpose library of works for various uses for which the company might
have of them, if any, was a different use altogether.

. . .

The downloaded pirated copies used to build a central library were not justified by a fair
use. Every factor points against fair use. Anthropic employees said copies of works (pirated
ones, too) would be retained “forever” for “general purpose” even after Anthropic determined they would never be used for training LLMs. A separate justification was required for each use. None is even offered here except for Anthropic’s pocketbook and convenience.

This Mintz blog has some thoughts on what the Bartz decision means for businesses using AI:

AI Training May Be Protected—But Not All Uses Are Equal –  If your business is training AI models, this ruling provides some legal cover—but only for the training process itself. Storing unauthorized content for other uses may not be permitted.

Expect More Litigation and Uncertainty – This is not the final word. The ruling is likely to be challenged, and future courts may interpret fair use differently. Companies should monitor developments closely and consider legal reviews of their AI data pipelines.

Audit Your AI Supply Chain – Companies using third-party AI tools should ask vendors where their training data comes from. If vendors can’t provide clear answers, that could be cause for concern.

Prepare for Contractual Shifts – We expect to see new indemnity clauses and representations and warranties in commercial contracts involving AI tools. Companies should be ready to negotiate these terms.