AI companies destroy physical books – let's scan rare books before it's too late

So here's the thing that gets under my skin about this whole AI training data gold rush: it's not just that companies like OpenAI and Anthropic are vacuuming up copyrighted works without permission—though that's bad enough. It's that the physical infrastructure of human knowledge itself is being treated as a disposable resource. Anna's Archive is pointing at something that should make anyone with a soul flinch: when AI companies buy up books to scan, they're often destroying the physical copies afterward. Maybe it's cheaper to shred them than store them. Maybe they don't want the provenance tracked. Either way, we're watching the raw material of civilization get incinerated in the name of "progress." The deeper irony here is that the very thing AI is supposed to preserve—knowledge, literature, the accumulated wisdom of generations—is being sacrificed to build models that will serve up a hollow, statistically plausible echo of it. Rare books are the archive of what we've thought, dreamed, and argued about for centuries. They're not just data points; they're physical artifacts of human struggle. And now they're being pulped to make way for a chatbot that will confidently tell you the wrong answer about them. I want to be clear: this isn't just a problem for copyright maximalists or library nerds. This is a civilizational issue. When you destroy the last copy of a book, you're not just losing a text—you're losing the annotations, the marginalia, the binding, the physical evidence of how people read and thought. No amount of OCR is going to capture that. The fact that this is happening in service of "AI advancement" makes it worse, because it frames the destruction as a necessary evil, a cost of doing business in the digital age. The call to action here is simple: scan rare books before it's too late. And honestly, that should be a no-brainer. We have the technology, the infrastructure, and the will—if we choose to exercise it. But it's going to take money, coordination, and a willingness to treat preservation as a public good, not a corporate afterthought. The alternative is a future where the only record of half the books humanity has ever written is a statistically probable string of tokens in a model's weights. And that's not a future I want to live in. Source: https://annas-archive.gl/blog/physical-destruction.html