AI labs are panic-buying paper books to escape web slop
Tech giants poisoned the internet with their own synthetic text and are now quietly buying up physical printed books to save AI from model collapse.
Book metadata platform ISBNdb has quietly pivoted into a bulk physical book broker, helping tech companies purchase between 1,000 and 1,000,000 physical volumes per order. Every procurement deal is signed under strict non-disclosure agreements to hide the tech labs buying up offline print stocks.
The sudden rush for real ink and paper stems from a self-inflicted crisis in AI engineering: the open web is now so saturated with AI-generated text that scraping fresh web data poisons new models with recursive synthetic garbage. Human-authored books published before 2022 have become the ultimate untainted data source required to prevent catastrophic model collapse.
Independent booksellers on marketplaces like Alibris and Biblio reported sudden automated sweeps clearing out hundreds of physical titles a week. Mike Feldman, customer service director at Alibris, confirmed that multiple wholesale accounts recently began buying up fiction inventory in massive automated batches.
This physical paper grab comes alongside severe copyright fallout, including Anthropic scanning and destroying millions of library books sourced from Better World Books, which led to a $1.5 billion court-approved copyright settlement with book publishers.
Silicon Valley burned through the entire digitized internet only to realize that dusty paper books saved on physical shelves were the only thing keeping artificial superintelligence from collapsing into complete gibberish.
Source: 404 Media
Comments
This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.