Technology and Science06 August, 20262 min read
Anthropic Bought Millions of Books for Destructively Scanning Under Project Panama
This article is an automated aggregation of coverage from news outlets. Our editorials offer in-depth analysis curated by the NewsCord team.
Anthropic bought millions of books, rare/high-quality titles, for destructively scanning under Project Panama. The purchases were global, with bulk orders from various countries to train AI models.
Project Panama and Books
A review of internal documents at Anthropic says the company bought millions of books in bulk to "destructively scan" them as part of a confidential effort known as "Project Panama." Snopes reported that a judge ordered the documents to be unsealed in the context of a copyright lawsuit, and that the documents show Project Panama was meant to scan and destroy "all the books in the world."
The Atlantic described how the internet raged after accusations that AI companies were "destroying the world’s books" by chopping off their spines to scan them and discarding them afterward. The Atlantic also said the practice intensified after an article by 404 Media suggested rare books might be included and named a book-database company called ISBNdb as the possible buyer, while noting that the 404 Media story acknowledged no evidence directly connected ISBNdb to the wave of purchases described by booksellers.
Court Findings and Denials
In the Bartz v. Anthropic dispute, a federal judge in the Northern District of California ruled the practice was "fair use" under the U.S. Copyright Act, with the logic that "Each copy printed was copied" and the original was destroyed.
The Week said court documents described Anthropic’s internal programme known as “Project Panama,” concluding books were “essential” for training advanced AI models, and it quoted a senior executive saying books taught systems “how to write well”.
Snopes reported that Anthropic denied buying and destroying “rare” or “antiquarian” books, with a spokesperson saying, "None of our data acquisition programs buy and destroy 'rare' or 'antiquarian' books,".
Snopes also quoted an internal memo showing the effort was intended to be kept secret, including the line: "We don't want it to be known that we are working on this."
What’s at Stake Next
The Atlantic said the outrage left unanswered questions about how rare the books being bought actually were and whether ISBNdb was responsible, while also stating that at least one major AI company used physical books to train its models. The Week warned that the trend has “raised fresh concerns about copyright and the future of physical books,” and it quoted Thomas Koch of the German Publishers and Booksellers Association saying it “appears to be yet another example of AI companies using vast quantities of copyright-protected works to train their language models, without consent and without payment.”
Snopes said it was not possible to confirm whether other AI companies have carried out similar projects, even as it described social-media claims on Instagram, Facebook, and Reddit that some rare or obscure titles were being scanned and destroyed. WWWhat's new | Other added that Ingram Content Group reacted by introducing an opt-out form for editors who do not want their books sold directly to AI companies for training, describing it as the first active consent mechanism in the sector.