A federal judge approved the largest copyright settlement in U.S. history on Monday, clearing Anthropic to pay roughly $3,000 per pirated work to authors and publishers while leaving the industry's underlying fair-use fight for AI training unresolved everywhere else.
For the data-hungry AI labs that have spent two years arguing over what counts as fair game for training, the math finally has a price tag. U.S. District Judge Araceli Martinez-Olguin granted final approval Monday of Anthropic’s $1.5 billion settlement resolving a class action brought by authors and publishers, according to TechCrunch. The deal, which covers more than 480,000 works at roughly $3,000 apiece, is described by the plaintiffs’ lead attorney, Justin Nelson, as “the largest known copyright recovery in history,” as reported by the New York Post.
The case, known as Bartz v. Anthropic, was filed in 2024 by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, who accused the AI company of downloading more than seven million pirated books to build Claude’s training library, according to 1310 WIBA. That number, and the roughly $3,000-per-work payout, matter beyond the courtroom: they set a rough price point for what unauthorized book scraping can cost, at a moment when publishers and authors elsewhere are trying to figure out what their own catalogs might be worth in litigation against other labs.
fair use, but not for piracy
The settlement exists because of a split ruling, not a clean win for either side. Judge William Alsup, who handled the case before retiring, found last year that training an AI model on copyrighted text counts as fair use — a decision TechCrunch called “widely seen as a turning point for the AI industry.” But Alsup drew a hard line between training on legally acquired books and building a library from pirate sites like Library Genesis and Pirate Library Mirror, ruling the latter illegal on its own terms and clearing the way for a trial on damages that could have run into the hundreds of billions of dollars, per the New York Post. Anthropic settled rather than risk a jury verdict.
who cashed in, who opted out
Claims covering more than 92% of the settlement’s roughly 480,000 works were filed, an attorney for the authors said during a court hearing, according to the Post. Martinez-Olguin overruled objections that the deal was too small, that it overcompensated plaintiffs’ attorneys, or that it wrongly excluded some rights holders, calling size complaints “not grounded in a realistic assessment of the overall risks and rewards of a trial.” She still awarded plaintiffs’ counsel more than $101 million of the $187.5 million they had requested. Some authors and publishers opted out entirely, judging the payout too low, and have filed separate suits against Anthropic that remain pending, according to Engadget. Under the terms, Anthropic must also destroy all original and duplicate files obtained from the pirated repositories — though the deal does not function as a license for future training or outputs, per 1310 WIBA.
no precedent, just a check
Because Anthropic settled instead of appealing, Alsup’s fair-use finding stays a single district court opinion — persuasive, perhaps, but not binding on any other judge. As TechCrunch put it, “Anthropic’s decision to settle means the case will never reach an appeals court to become binding precedent.” That leaves a crowded docket of unresolved fights against Google, Meta, Midjourney, and OpenAI over the legality of training on copyrighted works, each of which will be decided on its own facts by its own judge. Just last week, a group of publishers and authors — including Hachette, Cengage, Elsevier, novelist Scott Turow, and S.C.R.I.B.E. — filed a fresh class action against Google alleging it used their copyrighted works to train Gemini, per TechCrunch.
what to watch
For data companies and rights holders alike, Bartz v. Anthropic is now a benchmark rather than a rulebook: a $1.5 billion price tag and a $3,000-per-work figure that plaintiffs elsewhere will cite, but a legal theory — fair use for training, liability for piracy — that still has to be litigated fresh in every other courtroom. Watch the opt-out lawsuits against Anthropic, the newly filed Google case over Gemini, and whichever of the Meta, Midjourney, or OpenAI suits reaches a ruling first; any of them could either reinforce or scramble the fair-use line Alsup drew before this case ever got the chance to be appealed.
We are gratified by the Court's ruling granting final approval of this historic settlement. It is the largest known copyright recovery in history. We look forward to making distributions to the Class as promptly as possible.
