Anthropic's Book-Scanning 'Destruction' Draws Fresh Scrutiny
In one line: CounterPunch labels Anthropic's mass scanning and disposal of print books "Project Panama" and criticizes how the company sourced its AI training data.
Key points
- Anthropic reportedly bought print books in bulk to build a Claude training dataset, cut off their bindings to scan the pages, and then discarded the physical originals.
- CounterPunch frames the process as "Project Panama," objecting to the physical destruction of books on cultural and environmental grounds.
- The practice ties back to issues aired in 2025 copyright litigation (including Bartz v. Anthropic), where training on legally purchased, digitized books was treated as fair use while acquiring unauthorized copies was viewed separately.
Why it matters
It shows that how AI firms source training data is moving beyond legal rulings into ethical and cultural debate. Even "legally purchased" material leaves open the separate question of whether large-scale physical destruction is justified.
Read more
- Project Panama: Scanning and Book Destruction at Anthropic — CounterPunch
How this story unfolded
- LLM
- — Large Language Model의 약자로, '거대 언어 모델'이라고 해요. ChatGPT, Claude 같은 AI가 바로 LLM이에요. 엄청나게 많은 텍스트를 학습해서 사람처럼 글을 쓰고 대화할 수 있어요.