I wonder just how much of this "shadow library" content is stuff that's actually in the public domain and could be mirrored with no legal jeopardy whatsoever.

I wonder just how much of this "shadow library" content is stuff that's actually in the public domain and could be mirrored with no legal jeopardy whatsoever. Unfortunately, the low quality of catalog-like metadata that's available from so-called "shadow libraries" (including the one that's linked in the top comment) makes this a very hard question to answer. Even if that makes up only a handful of TB's or so, it would be worth mirroring the stuff - among other things, a reliable repository of copyright-free content would also be a valuable resource for "ethical" AI training and other such uses. (I know that the linked blogpost mentions that they just don't bother mirroring "widely available collections" of public domain books. The interesting question is whether there's some "long tail" of PD content that might not have made it to the more well-known collections as of yet.)