Project Panama: Anthropic’s destructive book scanning revealed
Confirmed
The takeaway
On Jan 27, 2026 the Washington Post reported, from court filings unsealed the week before in Bartz v.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 3 of 5
- Last verified
- 15 October 2026
Your AI and this story
- GPT-6 AstraIn its training data
- Claude Opus 5.5In its training data
- Gemini 4 ArgonIn its training data
- Grok 4.7In its training data
It happened before the cutoffs of all four assistants, so it can be in their training data.
Key facts
- Internal planning document: ‘Project Panama is our effort to destructively scan all the books in the world’ and ‘We don’t want it to be known that we are working on this’ (Washington Post via Boston Globe)
- Started early 2024; ‘tens of millions of dollars’ spent within about a year; Anthropic hired Tom Turvey, a Google executive who helped create Google Books (WaPo via Boston Globe)
- Sources: mainly used booksellers Better World Books and World of Books; a Strand spokesperson said the Strand never sold books to Anthropic; a vendor proposal aimed to convert ‘500,000 to two million books over a six-month period’; the number scanned is redacted (WaPo via Boston Globe)
- Filings also say Ben Mann downloaded books from LibGen over 11 days in June 2021, and in July 2022 sent colleagues a Pirate Library Mirror link with ‘just in time!!!’ (WaPo via Boston Globe; Euronews)
- Anthropic (deputy general counsel Aparna Sridhar): ‘This case has been settled, but the court’s landmark June 2025 ruling remains intact’; the settled issue was how some materials were acquired, not whether they could be used to develop AI models
- Exhibits are on the public docket, e.g. Dkt. 560-26, the internal ‘Project Panama’ page, filed publicly Jan 21, 2026 (CourtListener)
- Date caveat: Euronews (Aug 5, 2026) describes the documents as unsealed in late July 2026; the Washington Post story and the docket exhibits date the unsealing to January 2026
- Later finding: Japanese distributor Nippan confirmed in October 2026 that it sold books to Anthropic for the project (see the Nippan entry)
What happened
Exhibits from the authors’ class action, filed publicly in January 2026, showed how Anthropic built a book corpus legally after its earlier use of pirated libraries: buy physical books, cut them apart and scan them. Judge Alsup had ruled in June 2025 that scanning lawfully bought print books for training was fair use, which is why Anthropic presents the settlement as being only about how some material was acquired. The Washington Post article itself is paywalled; this entry uses the Boston Globe’s syndicated copy.
Why it matters
Project Panama shows the scale of the race for high-quality text and the trade-off Anthropic made (destructive scanning was cheaper and faster than Google Books’ non-destructive method). It resurfaced in coverage through 2026, fed the music publishers’ August 2026 suit, and led to vendor disclosures such as Nippan’s.
Sources
4 sources from 4 sites. Numbers match the chips in the text.
4 sources: 1 primary, 3 press
Primary
- Court exhibit: ‘Project Panama’ internal page (Bartz v. Anthropic, Dkt. 560-26)storage.courtlistener.com, docs
Press
- Washington Post: Anthropic ‘destructively’ scanned millions of books to build Claude (Jan 27, 2026; paywalled)washingtonpost.com, press
- Boston Globe (Washington Post syndication): Silicon Valley’s plan to destroy booksbostonglobe.com, press
- Euronews: Project Panama, how Anthropic secretly destroyed millions of books to train its AI (Aug 5, 2026)euronews.com, press
Changes
- Filed (WaPo read through the Boston Globe syndication)