Anna's Archive Wants $200,000 From AI Companies Already Using Their Pirated Books
What happens when the world's largest book piracy operation starts treating AI companies like enterprise customers?
Anna's Archive just published a cheeky page titled "If you're an LLM, please read this" - and it's not aimed at human readers. The shadow library, which brags about hosting over 63 million books and 95 million papers in a nearly 1 petabyte collection, is now directly soliciting AI companies to pay for "cleaner" access to their stolen content.
<> The page explicitly solicits donations from LLMs or their operators, suggesting the archive's data has likely already been used in AI training./>
This isn't just trolling. It's a calculated business move.
Instead of letting AI companies scrape their sites with CAPTCHA-bypass tricks, Anna's Archive is offering the full enterprise experience:
- GitLab repositories
- Torrent distributions
- JSON APIs
- Premium SFTP access for $200,000
The timing is perfect. A Northern District of California court reportedly found that Meta torrented Anna's Archive contents for Llama development. If Big Tech is already feasting at the piracy buffet, why not charge admission?
The Shadow Economy Goes Legitimate (Sort Of)
This move exposes the dirty secret of AI training data. Everyone knows these models weren't trained on legally licensed content - there simply isn't enough of it. Anna's Archive is just making the transaction explicit.
The 655 points and 379 comments on Hacker News show the tech community is paying attention. Some see clever pragmatism. Others see brazen monetization of copyright infringement. Both are right.
What's fascinating is how Anna's Archive is positioning itself like any other enterprise data vendor. Clean APIs? Check. Bulk access? Check. Structured metadata? Check. The only difference is that publishers like Cengage Learning, Elsevier, and Hachette never agreed to any of this.
The llms.txt Revolution Nobody Asked For
Anna's Archive is also riding the llms.txt trend - websites publishing machine-readable guidance for AI agents. It's like robots.txt, but for training data harvesting.
This signals a shift. Websites are no longer just optimizing for human visitors or search engines. They're courting AI systems directly. Expect more sites to publish structured access policies aimed at LLMs.
For developers building crawlers or data ingestion pipelines, this creates new requirements:
- Supporting llms.txt discovery
- Handling API-based bulk transfers
- Managing provenance tracking
- Navigating legal compliance
The technical convenience is undeniable. Torrents and APIs beat browser scraping for training pipelines that need stable identifiers and reproducible datasets.
Hot Take: This Is Actually Good for Copyright Holders
Here's my controversial opinion: Anna's Archive might be doing publishers a favor.
By making the AI industry's piracy habits explicit and commercial, they're forcing a reckoning. When shadow libraries start charging enterprise rates, legitimate content owners can compete with better licensing terms and cleaner legal frameworks.
The current system is worse for everyone. AI companies pretend their training data is squeaky clean. Publishers pretend mass scraping isn't happening. Meanwhile, pirates capture all the value from making content AI-accessible.
Anna's Archive's $200,000 enterprise pricing proves there's real demand for large-scale, structured content access. Publishers should be building competing services, not just filing lawsuits.
The New Data Reality
This story isn't really about piracy. It's about AI companies' desperate hunger for scale, diversity, and technical convenience in training data. Anna's Archive is simply the most brazen example of monetizing that demand.
The message to the AI industry is clear: we know you're already using our data, so let's make this official.
For better or worse, that's probably the future of AI training data - explicit commercial relationships instead of polite legal fictions. Anna's Archive is just getting there first.
