$1.5B Twist: Pirated Books, but AI Training Was Legal
Anthropic was ordered to pay $1.5 billion for pirated books, yet a judge ruled AI training lawful, exposing a copyright brawl under 1976 laws still on the books.
Summary
Anthropic got hit with a $1.5 billion settlement for pirated books, but its AI training itself was ruled lawful. In 2025, Judge William Alsup said training models on copyrighted writing can resemble reading, not copying, even as Anthropic projects about $200 billion in annual revenue by 2028.
That split defines the mess around ChatGPT, Gemini and Claude. The 1976 copyright law leaves courts wrestling with fair use. Judge Stephanos Bibas ruled Ross Intelligence could not use Thomson Reuters content to build a direct rival. Thaler v. Perlmutter found fully AI-generated work cannot be copyrighted. Attorneys Cathy Gellis and Jason Henderson say pending lawsuits mean no clean answer is coming soon.
Positives
- Judge William Alsup ruled Anthropic’s AI training itself was lawful.
- Fair use can protect transformative training that does not directly compete with the original.
- The rulings give AI companies early guidance while bigger lawsuits continue.
Risks & concerns
- Anthropic was ordered to pay writers a mammoth $1.5 billion over pirated books.
- Ross Intelligence lost after using Thomson Reuters content to build a direct competitor.
- Conflicting rulings and a 1976 law leave authors and AI companies in limbo.