Project Panama: the revelation of thousands of books destroyed to train AI

A secret project shook the publishing industry and artificial intelligence. Leaked documents from a lawsuit against Anthropic revealed “Project Panama”, a confidential initiative that acquired and destroyed thousands of used books to feed AI models with their texts.

How it worked

Sellers of old and used books received massive purchase orders for titles. The books were cut with hydraulic guillotines — physically destroyed — in order to scan them at high speed and turn their content into training data. The industry’s fear is that thousands of works are lost forever.

The underlying debate

The revelation revives questions about copyright in the AI era: can companies use protected works without permission? Is physically destroying books worth it to accelerate technological development? The case promises to set precedent on model training.

The key figure

Thousands of destroyed books. The cultural cost of artificial intelligence’s insatiable appetite for data.