TECH Signal 442
Advocacy groups urge FTC to probe AI firms' book-scan-and-destroy practices
↗
Advocacy groups have asked the FTC to investigate AI companies that scan physical books for training data and then destroy the originals, citing anticompetitive and knowledge-hoarding concerns.
Engineers building or training models may face legal scrutiny if book-scan-and-destroy practices are deemed anticompetitive or unlawful, potentially limiting a low-cost data source. The controversy also highlights reputational risks and could push firms toward more expensive or less effective data acquisition strategies, affecting project timelines and budgets.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Advocacy groups claim AI firms' book-scan-and-destroy operations hoard knowledge and hinder competition.
Internal documents show Anthropic called its effort "Project Panama" and aimed to destructively scan all books.
Amazon and intermediary ISBNdb have also been linked to book scanning and shredding activities.
THE READ
What the cluster adds up to.
The reported change involves AI companies purchasing large quantities of physical books, scanning the text for model training, and then destroying the original copies. Advocacy organizations have submitted a letter to the US Federal Trade Commission asking for an investigation into these book-buy-and-destroy operations, arguing they constitute knowledge hoarding and anticompetitive behavior.
Internal communications reveal that Anthropic referred to its initiative as "Project Panama" and described it in a 2024 memo as "our effort to destructively scan all the books in the world." The company reportedly chose the codename to keep the activity confidential. Older books are prized for training because they lack AI-generated text, and destroying them avoids ongoing storage expenses.
Critics contend that the practice removes non-renewable works from public access, potentially allowing only well-funded incumbents to build high-quality models while restricting competitors and the broader public. The advocacy letter warns that such actions could engineer a future where a few firms control humanity's written works after having eliminated the originals.
Legal defenses based on fair use have had mixed results; a district court accepted the argument in the Anthropic case but rejected it for the Internet Archive. If the FTC pursues action, firms may face restrictions or penalties that force them to retain scanned books, seek alternative data sources, or incur higher costs, thereby altering the economics of model training.
Written by elseif from the cluster below · checked for specifics the sources never containedTHE CLUSTER