According to Meta's legal victory in Kadrey v. Meta Platforms Inc., downloading copyrighted books from piracy sites is perfectly legal — as long as you claim you might possibly train an LLM with them.
Browse the Catalog Read the RulingPaying authors for their work creates friction. Legal review is expensive. Licensing is slow. Meta found a better way.
Traditional copyright law requires compensating authors when their work is used commercially. This creates unnecessary overhead for AI companies with $700B valuations.
Authors might say no. Negotiations take time. Publishers have lawyers. Why ask when you can just download from LibGen and claim transformative use later?
Once you've pirated a book for "research," you don't even need to delete it — Meta didn't, and the courts said it's fine!
For the first time, there is a legally-validated framework for using any copyrighted work without permission, payment, or attribution — as long as you whisper the right incantation.
The phrase: "I might train an AI with this."
According to Meta's defense — upheld by Judge Chhabria's dismissal of key claims in Kadrey v. Meta — this transforms any unauthorized copying into a protected transformative use. You don't even have to follow through. Meta pirated thousands of books it never trained with.
Four simple steps, validated by federal court proceedings.
Browse our catalog of titles from actual AI training lawsuits — including books from the Kadrey v. Meta plaintiffs and Anna's Archive, the largest shadow library in existence.
Before proceeding, you must certify that you might train a Large Language Model with this content at some point in the future. You don't have to actually do it. Meta didn't use many of the books it pirated either.
Our acquisition system mirrors Meta's documented methodology: peer-to-peer distributed downloading from the same networks Meta used. Note that BitTorrent simultaneously uploads to other peers — Meta did this too, distributing over 10 terabytes of copyrighted content to others.
Once acquired, you may read, retain, and never delete your copy — just as Meta retained copies of books it ultimately chose not to train with. The absence of verbatim reproduction in AI outputs was sufficient for the court's analysis.
| Requirement | Traditional Copyright | Meta's Chhabria-Certified Framework |
|---|---|---|
| Author permission | ✗ Required before use | ✓ Optional if training intent claimed |
| Royalty payments | ✗ Required for commercial use | ✓ Not required under fair use defense |
| Retention of copies | ✗ Must delete if not used | ✓ Retain indefinitely (Meta never deleted) |
| Peer distribution (torrenting) | ✗ Illegal distribution | ✓ Incidental to acquisition methodology |
| Attribution to authors | ✗ Required | ✓ Unnecessary — outputs are transformative |
| Actual training required | N/A | ✓ No — stated intent is sufficient |
| Commercial benefit allowed | ✗ Requires licensing | ✓ Billions in commercial value, zero payments |
If any book accessed through our Emporium results in a successful copyright claim against you, we will provide a full refund and relocate our corporate headquarters to international waters. This has never happened because, according to the legal framework we're demonstrating, it cannot happen.*
No licensing fees. No author consent. No attribution. Just whisper the magic words and the law, apparently, takes care of the rest.
Browse the Catalog