A federal judge approved the largest AI copyright settlement in US history on July 20, 2026, awarding $1.5 billion to Anthropic for using 482,460 copyrighted books in training data. But the case is far from over. Some 350 authors who rejected the settlement are now litigating separately, potentially exposing Anthropic to far larger payouts under federal copyright law.

What Happened in the Bartz Settlement?

Judge Araceli Martínez-Olguín granted final approval to the settlement after overruling all 53 objections filed by class members. The deal covers a certified list of 482,460 books, which works out to roughly $3,109 per work on average. The settlement also included approval of the release scope, meaning Anthropic received broad protection from future claims related to the books used in its training process.

The settlement efficiency ratio tells an important story about how far apart the two sides were. Axis Intelligence Research calculated that the $1.5 billion represents just 2.07 percent of the theoretical maximum damages available under federal copyright law, which could have reached $72.369 billion under 17 U.S.C. Section 504(c). In other words, Anthropic negotiated a settlement worth roughly one-fiftieth of what a jury might have awarded if the case had gone to trial.

Why Are 350 Authors Still Suing?

Not every author accepted the settlement. The 350 who opted out are now pursuing separate litigation, and they have a powerful incentive to do so. Federal copyright law allows statutory damages of up to $150,000 per infringed work, regardless of actual harm. For the 1,802 works these opt-out plaintiffs claim were used without permission, the gap between what they received in the settlement and what they could win in court is substantial.

Axis Intelligence Research quantified this gap as an "Opt-Out Exposure Delta" of $264.7 million. That figure represents the difference between the statutory ceiling available to opt-out claimants and the value those same works received inside the class settlement. In practical terms, some authors believe they can recover far more by litigating individually than they would have received as part of the class.

What Does This Mean for the Broader AI Copyright Landscape?

The Bartz settlement is closed, but the legal landscape for AI training and copyright remains unsettled. As of September 7, 2026, no US appellate court has ruled on whether using copyrighted works to train AI models qualifies as fair use. The Third Circuit heard oral arguments on June 11 in Thomson Reuters v. Ross Intelligence, a case that could become the first appellate decision on AI training fair use, but thirteen weeks later, no ruling has been issued.

Meanwhile, courts around the world are reaching different conclusions. On July 31, 2026, the Munich District Court in Germany ruled in GEMA v. Suno that US copyright law applies to training conducted in the United States and rejected the fair use defense. That decision suggests international courts may take a harder line on AI training than US courts have so far.

How to Track Active AI Copyright Cases

• Settlement Status: The Bartz v. Anthropic case is concluded with final approval granted, but opt-out litigation continues separately with no scheduled trial date yet announced.
• Appellate Rulings Pending: The Third Circuit has not yet ruled on fair use in Thomson Reuters v. Ross Intelligence, which could set precedent for all future AI training cases in the United States.
• International Developments: German courts have rejected fair use defenses for AI training, and the European Court of Justice is considering a reference case (Like Company v. Google Ireland) that could reshape AI copyright law across the EU.
• Music Industry Litigation: Multiple cases involving music training data remain active, including Sony Music v. Suno and UMG v. Suno, with summary judgment motions expected in 2027.
• Visual Content Cases: Lawsuits against Midjourney, Stability AI, and DeviantArt over image generation remain in discovery, with trial dates and class certification schedules still being determined.

The settlement landscape is evolving rapidly. As of September 2026, Axis Intelligence Research tracks 46 active AI copyright cases across the United States and internationally, with major developments updated weekly and verdicts or settlements exceeding $500 million reported within 24 hours.

For authors, artists, and publishers, the Bartz settlement offers a cautionary lesson. The $1.5 billion payout sounds large in absolute terms, but it represents a fraction of what copyright law technically allows. The decision by 350 authors to opt out and pursue separate claims suggests that some rights holders believe they can do better by taking their chances in court rather than accepting a negotiated settlement.

The broader question remains unanswered: Is using copyrighted works to train AI models fair use? Until an appellate court rules on that question, both AI companies and copyright holders will continue to operate in legal uncertainty. The Bartz settlement may be closed, but the war over AI copyright is far from over.