A reporter at 404 Media hid an AirTag in a rare book, watched where it went, and traced it to an Amazon warehouse in Las Vegas. There, a team strips the bindings off books so they can be scanned faster, destroying the physical copy. The scanned text becomes training material for AI models.
The part worth your attention is why the books are valuable. AI companies have already scraped almost everything freely available on the open web. Printed books contain text that was never online, and anything printed before 2022 is guaranteed not to contain AI-written filler, which degrades models trained on it.
Read that back with your own website in mind. If a company is willing to buy and destroy physical books to find text that hasn't been scraped, your site's pages have already been read, indexed and absorbed. Your service descriptions, your FAQ, your blog posts, your about page. That happened without anyone asking you, and there is no undo button.
What this changes for you day to day: almost nothing. Your site works the same tomorrow. Your customers won't notice. There's no security issue and no update to install.
What it does change is how you think about the words on your site. Two practical consequences. First, generic copy that reads like everything else on the internet is now genuinely worthless as a differentiator, because a chatbot can produce it instantly. Specific detail (your actual process, your actual pricing structure, the questions your actual customers ask) is the only thing that can't be synthesised. Second, if you publish something you consider proprietary, a training manual, a detailed methodology, a pricing calculator, assume it will be ingested. Put it behind a login or a form if that matters to you.
What to do: Nothing urgent. Next time you speak to whoever maintains your site, ask two things. Whether your robots.txt file (a plain text file that tells automated crawlers what they may and may not read) blocks the known AI crawlers, and whether anything genuinely proprietary is sitting on a public page where it doesn't need to be. Neither is an emergency. Both are ten-minute conversations.
Reported by:
