Anthropic has admitted to using millions of pirated books and destroying them to train its AI model, Claude. The practice, revealed in a class-action lawsuit, has sparked controversy and legal scrutiny, with a $1.5 billion copyright settlement recently approved.
TL;DR
- Anthropic used millions of pirated books and destroyed them to train its AI model, Claude.
- The practice was revealed in a class-action lawsuit, with a $1.5 billion copyright settlement recently approved.
- The controversy highlights the ethical and legal challenges AI companies face in sourcing training data.
What happened
Anthropic, the company behind the AI model Claude, admitted to using millions of pirated books to train its model. The practice, known as Project Panama, involved purchasing print books, stripping them from their bindings, scanning them, and discarding the originals. This was revealed in a class-action lawsuit filed by the Authors Guild, which resulted in a record $1.5 billion copyright settlement approved in June 2025.
The lawsuit also revealed that Anthropic considered using pirated material from online databases but decided against it. Instead, the company opted for a physical book-destruction operation, believing it to be the most cost-effective means to achieve a world-class LLM. According to court documents, Anthropic sought to convert between 500,000 to two million books over a six-month period.
Why it matters
The controversy surrounding Anthropic's book destruction practice highlights the ethical and legal challenges AI companies face in sourcing training data. While the practice may be legally permissible under the definition of fair use, it has sparked public outcry and negative PR for the company. The controversy also raises questions about the environmental impact of such practices and the potential loss of cultural heritage.
For developers and startups, this controversy serves as a cautionary tale about the importance of ethical data sourcing and the potential legal and reputational risks of cutting corners. For investors, it underscores the need for due diligence and transparency in AI companies' data practices.
Key facts
- Anthropic used millions of pirated books to train its AI model, Claude.
- The company destroyed the books after scanning them, a practice known as Project Panama.
- A class-action lawsuit filed by the Authors Guild resulted in a record $1.5 billion copyright settlement approved in June 2025.
- Anthropic considered using pirated material from online databases but decided against it.
- The company sought to convert between 500,000 to two million books over a six-month period.
- The controversy has sparked public outcry and negative PR for Anthropic.
- The practice raises questions about the ethical and legal challenges of AI companies' data sourcing practices.
Context
The controversy surrounding Anthropic's book destruction practice is not an isolated incident. Other AI companies have also been caught slicing up old books to create training data for their LLMs. This practice has sparked a debate about the ethical and legal implications of AI companies' data sourcing practices.
The controversy also highlights the growing tension between AI companies and creators, with AI companies often being accused of using creators' work without permission or compensation. This tension has led to calls for greater regulation and oversight of AI companies' data practices.
As AI continues to evolve and become more integrated into our daily lives, the ethical and legal challenges of AI companies' data sourcing practices will only become more pressing. It is crucial for developers, startups, and investors to be aware of these challenges and to advocate for ethical and transparent data practices.
