Anthropic settled a copyright lawsuit for $1.5 billion, the largest in U.S. history, despite a judge ruling that training AI on books was fair use. The payment was for obtaining the books unlawfully, not the training itself.
TL;DR
- Anthropic paid a record $1.5 billion for using pirated books to train its Claude AI model, despite a judge ruling the training itself as fair use.
- The settlement highlights the importance of data provenance in AI development, not just the transformative use of the data.
- Small businesses using AI should scrutinize their data sources and terms of service to avoid similar legal pitfalls.
What happened
In Bartz v. Anthropic, Judge William Alsup ruled that training AI on copyrighted books was fair use under Section 107 of the Copyright Act. He described the technology as 'among the most transformative many of us will see in our lifetimes.'
Despite this ruling, Anthropic agreed to pay a record $1.5 billion settlement, the largest in U.S. history, according to the Authors Guild. The settlement was for the way the company obtained the books, not the training process itself.
Judge Alsup found that while training the AI was transformative, building a permanent library of pirated copies was not. Anthropic had the means to purchase the books but chose to steal them to avoid 'legal/practice/business slog,' as stated by co-founder and CEO Dario Amodei.
Why it matters
This case underscores that the provenance of data is crucial in AI development. Businesses must ensure they have lawful access to the data they use, not just that the use is transformative.
For small businesses, this ruling serves as a warning to scrutinize their data sources and the terms of service of AI tools they use. Unauthorized use of copyrighted material can lead to significant legal and financial repercussions.
The settlement may encourage more licensing deals between AI companies and content creators, as suggested by Mary Rasenberger, CEO of the Authors Guild. This could lead to a more structured market for AI training data.
Key facts
- Anthropic agreed to pay a $1.5 billion settlement, the largest copyright settlement in U.S. history, according to the Authors Guild.
- The settlement was for using pirated books to train the Claude AI model, not the training process itself.
- Judge William Alsup ruled that training AI on copyrighted books was fair use under Section 107 of the Copyright Act.
- Anthropic had the means to purchase the books but chose to steal them, according to Judge Alsup.
- The settlement works out to roughly $3,000 across an estimated 500,000 works, per the Authors Guild.
- Verizon's 2026 Data Breach Investigations Report found that 45% of employees regularly use AI on company devices, with 67% accessing AI services through non-corporate accounts.
- The Copyright Office has held that purely AI-generated work cannot be registered, as copyright requires human authorship.
Context
This case is part of a broader conversation about the economic implications of AI. James Douglas Cardoza, founder of LibertyPen, argues that AI raises centuries-old questions about ownership of an individual's labor.
The AI industry is facing roughly 120 copyright suits, with several cases potentially producing rulings that contradict the current one. The Copyright Office's position on training is guidance, not law, and has drawn industry pushback.
The outcome of these legal battles will shape the AI landscape, determining how companies access and use data, and how content creators are compensated.
