Anthropic Destructively Scanned Millions of Books to Legally Avoid Piracy Claims
soulbeddu · reddit · 2026-07-30
Recently resurfaced court documents reveal that Anthropic bought and 'destructively scanned' millions of physical books—cutting off bindings and discarding the copies—to build its training corpus. While some frame this as 'AI book burning,' the underlying legal logic is highly counterintuitive.
Destroying legally purchased books was partly a strategy to comply with 'format shifting' legal definitions, thereby avoiding the severe risks associated with downloading pirated files. In Bartz v. Anthropic, Judge William Alsup ruled that training on legally purchased, destructively scanned books constituted fair use, whereas using pirated books did not. Anthropic later agreed to pay roughly $1.5 billion to settle the piracy claims.
The debate now centers on the ethics of this practice: while acceptable for bulk paperbacks where text survives elsewhere, shredding rare or out-of-print editions results in the irreversible loss of the physical artifact.
More from Safety
- Agent actions need checks outside the model: NVIDIA demos lean on guardrails amid wave of real agent incidents — Creamy-And-Crowded · 2026-10-16
- Vitalik builds "paranoid mode" agent wrapper with Tor, zkAPI and local prompt rewriting — DavideCrapis · 2026-10-16
- Security team demos Signal 1-click exploit technique at Hexacon Paris — moyix · 2026-10-16
- China moves to impose registration and licensing on frontier AI model weights and chip designs — deanwball · 2026-10-16
- Same agent task, two auth models: shared admin login wipes 1M audit rows, per-agent identity gets refused — Prompt Engineering · 2026-10-16
- Binding agent approvals to exact content: a synthetic test for stale permissions — Crescitaly · 2026-10-16