When AI Learns From Books It Didn’t Own

TDT | Manama

Email: mail@newsofbahrain.com

Artificial intelligence can write, summarise and create in seconds — but one question remains difficult: where does AI get the knowledge it learns from, and who owns it?

That question became the centre of a landmark case involving AI company Anthropic, the maker of Claude. Authors accused the company of using copyrighted books without permission to train its AI systems. Court proceedings revealed that Anthropic had obtained millions of books, including material from unauthorised sources, as it built its training datasets.

One of the most striking parts of the case was Project Panama, an effort in which physical books were purchased, their spines removed, and their pages scanned for digital use. The physical copies were then discarded or recycled.

In July 2026, a US federal judge gave final approval to a $1.5 billion settlement between Anthropic and authors and publishers. It became the largest known US copyright settlement, with compensation calculated on eligible works.

The case is important because it separates two questions: can AI learn from copyrighted material, and can companies obtain that material illegally? A previous court ruling found that training on books could qualify as fair use, while Anthropic’s storage of pirated copies created a separate copyright problem.