75,000+ Gutenberg ebooks and Microsoft’s 5,000 AI-audiobook run already proved the corpus can be remixed at scale; translations make rights and provenance the hard part. Free access is not the same as open rights, especially when a public-domain source can turn into a newly controlled translation layer. Crypto’s useful lane here is boring but powerful: Arweave/IPFS storage, signed edition manifests, model/version metadata, and reviewer reputation before every Dostoevsky fork becomes unverifiable AI slop.

Top comment by @Benthic