- What: Judge Vince Chhabria issued a strategic order in Kadrey v. Meta Platforms challenging the fair use defense.
- When: The order was issued on March 25, 2026.
- The Conflict: Meta's use of copyrighted materials for LLM training faces a renewed "market harm" legal threat.
- Impact: Future plaintiffs may now bypass previous summary judgments to sue AI developers individually.
In a move that could dismantle the "fair use" legal shield currently protecting generative AI developers, Judge Vince Chhabria of the Northern District of California has signaled that the fight over copyrighted training data is far from over. In a new order issued March 25, 2026, Chhabria suggested that while Meta Platforms had previously secured a summary judgment in the Kadrey v. Meta case, that victory does not protect the company from a future wave of individualized lawsuits based on "market harm" theories.
The ruling marks a significant escalation in the judicial tug-of-war over how large language models (LLMs) are trained. It effectively breathes new life into claims that using copyrighted books and articles to train AI constitutes infringement, provided future plaintiffs can prove that such training directly damages the market for their original works.
The Resurrection of the Market Harm Theory
The legal battle over AI training reached a fever pitch in June 2025, when two conflicting rulings emerged from the Northern District of California. On June 23, 2025, Judge William Alsup ruled in Bartz v. Anthropic that using copyrighted materials to train LLMs was protected under the fair use doctrine, a decision that was widely celebrated by the AI industry.
However, just two days later, on June 25, 2025, Judge Chhabria issued a more critical ruling in Kadrey v. Meta. While he ultimately granted Meta's motion for summary judgment on the training claims, he did so with admitted reluctance. According to court observers and the recent order, Chhabria believed the plaintiffs' counsel had failed to properly argue the "market harm" factor of the fair use test—a failure he attributed to the legal team's inability to grasp the core economic impact of generative AI.
Under U.S. copyright law, "market harm" is the fourth and often most critical factor in determining fair use. It asks whether the new use of a work impairs the market for the original. Chhabria has long contended that training an AI on an author's work to create a tool that could eventually compete with that author is a clear form of market harm, even if the legal teams in front of him initially failed to prove it.
The "Subversive" Footnote
The latest development comes via a footnote in an order permitting a minor amendment to the remaining portions of the Kadrey complaint. While the primary case against Meta has slowed, Chhabria used the footnote to plant what legal wonks are calling a "legal time bomb" for the AI industry.
In the footnote, Judge Chhabria wrote:
"It seems far less likely that absent class members would be precluded from subsequently bringing training claims... The training claim will always be subject to a fair use defense. And the most important of the fair use factors—market harm—will often be highly fact-dependent, such that training claims would likely be individualized and therefore not precluded by a judgment against the class."
This statement is a direct invitation to other authors and copyright holders. It suggests that even though Meta "won" against the current named plaintiffs, the thousands of other authors whose books were used in Meta’s training sets are not bound by that loss. Because "market harm" varies from author to author, each could potentially bring their own suit, arguing that Meta’s AI specifically damages the market for their specific genre or style.
Impact on Meta, Anthropic, and the AI Industry
This shift from broad class-action suits to "individualized" claims represents a nightmare scenario for companies like Meta and Anthropic. If Chhabria’s theory holds, AI developers could face a "death by a thousand cuts" litigation strategy.
For developers, this means the legal certainty provided by early fair use wins is evaporating. If market harm is "highly fact-dependent," as Chhabria asserts, AI companies cannot rely on a single landmark ruling to protect their entire training pipeline. Instead, they may have to prove fair use over and over again for different categories of data.
For the broader industry, this signals a power shift back toward content creators. The "move fast and break things" approach to data scraping is hitting a judicial wall where the economic survival of the creator is being prioritized over the technical progress of the model.
What's Next: A Warning to Creators
Judge Chhabria’s order concluded with a subtle but urgent warning regarding the statute of limitations. By signaling that the door is still open for new claims, he also implied that it will not stay open forever.
"Do something!" seems to be the underlying message to the putative class of authors. The industry is now watching to see if new legal counsel—those whom Chhabria might view as more competent than the "lunkheads" he previously criticized—will step forward to launch a fresh assault on Meta’s training practices using the market harm roadmap the judge has provided.
As the Kadrey case limps forward, the AI industry must now contend with the reality that a single footnote in a California courtroom may have just reset the clock on the most existential legal threat to generative AI.