Anthropic’s landmark $1.5B copyright settlement is approved
A federal judge has given final approval to Anthropic's $1.5 billion settlement of a class action copyright lawsuit brought by authors and book publishers, clearing the way for payouts to begin. The case centred on Anthropic's practice of downloading and storing millions of copyrighted books to train its AI models, and the settlement is believed to be the largest in the history of US copyright law, marking a significant moment for the wider AI industry as it grapples with similar legal challenges elsewhere.
Under the deal, rights holders will receive $3,000 per work across an estimated 500,000 works. Judge William Alsup, who has since retired, had ruled that training AI on copyrighted text constitutes fair use, but found that Anthropic's use of pirated sources such as Library Genesis was unlawful, prompting the settlement rather than a trial; Judge Araceli Martinez-Olguin has now signed off on the final approval. Because the case was settled rather than appealed, the ruling sets no binding industry-wide precedent, and similar copyright disputes continue against companies including Google, Meta, Midjourney and OpenAI, with publishers recently suing Google over its Gemini AI training practices.
- Anthropic's $1.5bn book copyright settlement wins final court approval
- Authors and publishers to receive $3,000 per work, roughly 500,000 works
- Wider AI copyright legal battles continue against Google, Meta, OpenAI, Midjourney
New here? Start with this
Anthropic is a major artificial intelligence company that builds AI models similar to those made by OpenAI or Google. Like other AI developers, it trained its systems by feeding them huge amounts of text, and part of that material reportedly included millions of copyrighted books obtained without permission from publishers or authors. A group of authors and publishers took the company to court over this, arguing it amounted to copyright infringement.
Rather than go to trial, Anthropic agreed to pay $1.5 billion to settle the claims, with individual rights holders set to receive a fixed sum for each affected work. A federal court has now given final sign-off on this settlement, allowing payments to actually go ahead. The case is notable because of its sheer scale, and because a judge had earlier drawn a distinction between AI training itself, which was found lawful, and the use of pirated copies to do it, which was not.
The outcome matters beyond this one company because it feeds into a much wider, unresolved debate over how copyright law applies to AI training. Since Anthropic settled instead of appealing, the case does not set a binding legal precedent for others, and similar disputes are still working their way through the courts against firms including Google, Meta, OpenAI and Midjourney.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Supporters of the settlement and of the broader fair-use framework argue that training AI models on text, including copyrighted books, is a transformative process akin to how human writers learn from wide reading, and that Judge Alsup's underlying reasoning correctly separates this legitimate technological progress from the narrower, genuinely unlawful act of sourcing texts from pirated repositories. They view the deal as a proportionate resolution that compensates rights holders for the specific harm caused by piracy, while avoiding a sweeping ruling that could criminalise AI training generally and chill innovation and investment in an industry still finding its legal footing. For them, the size of the payout demonstrates that courts and companies alike take copyright seriously without requiring every act of machine learning to be treated as infringement.
The case against
Authors, publishers, and their advocates argue that $3,000 per work is a modest sum that fails to reflect the true commercial value AI firms have extracted from a lifetime of creative labour, especially where the resulting models can generate competing content at scale and potentially erode authors' future earnings. They contend that, because the case settled rather than went to trial or appeal, it leaves no binding precedent to deter future misconduct, effectively allowing large AI companies to treat copyright infringement as a manageable cost of doing business rather than an obligation to negotiate proper licences upfront. For these advocates, genuine respect for creators' rights requires consent and fair payment before their work is used, not compensation only after unlawful copying is uncovered.