Largest Pirate Library | U.S. Court Orders Anna’s Archive to Pay $19.5 Million, Global Domain Block, Is AI Trained on It in Trouble?

Southern District of New York Federal Court issues a default judgment against Shadow Library Anna’s Archive, awarding $19.5 million in damages along with a permanent global domain name ban; publishers also accuse Anna’s Archive of being a primary training data source for AI companies like Meta and NVIDIA.
(Background summary: SpaceX IPO filings reveal holdings of 18.7k Bitcoin! Average price $35,320, set to become the seventh-largest Bitcoin company worldwide in June)
(Additional context: Nvidia’s Q1 earnings blowout! Revenue hits $81.6 billion, a record high, Jensen Huang exclaims “The era of Agentic AI has arrived,” dividends surge 24 times)

Table of Contents

Toggle

• The $19.5 million on paper, versus the $322 million that can’t be recovered
• The real targets: Cloudflare, domain registrars, and offshore intermediaries
• The new battleground for AI training data

The $19.5 million and the global domain ban, signed by Judge Jed S. Rakoff of the Southern District of New York on the 19th, put a dollar figure on the legal cost for Shadow Library Anna’s Archive, but the case is more complex than it seems.

The $19.5 million on paper, versus the $322 million that can’t be recovered

Judge Rakoff awarded each of the 130 “litigation works” the statutory maximum of $150k in damages, totaling $19.5 million. U.S. copyright law allows up to $150k per infringed work, and Rakoff went all out.

However, whether the amount can actually be collected is another matter. Anna’s Archive operators remain strictly anonymous, having publicly stated that hiding their identity is to avoid “decades of imprisonment.” The default judgment requires them to disclose their identity within 10 days, but it’s almost certain they will ignore it.

In the music industry, in cases related to Spotify, Anna’s Archive has already been awarded a default judgment of up to $322 million, with no signs of actually recovering a single dollar.

The two figures, $19.5 million and $322 million, share a common point: they are “paper victories,” with real enforcement coming from elsewhere.

The real targets: Cloudflare, domain registrars, and offshore intermediaries

Publishers have long seen through Anna’s Archive’s tactic of “switching to backup domains when blocked,” so this time, the core weapon of the judgment isn’t damages but a permanent injunction: requiring “all domain registration authorities and registrars” to permanently disable its domains and prevent transfer, effectively freezing the domains’ continued circulation within the global DNS.

The court’s ruling names over 20 companies and organizations, including:

• Cloudflare (CDN and DDoS protection provider)
• Njalla (anonymous domain proxy service)
• DDoS-Guard (traffic protection service)
• TELE Greenland / Tusass (manages .gl domains)
• PKNIC (manages .pk domains)
• Grenada National Telecommunications Regulatory Authority (manages .gd domains)

But this list reveals a key gap: U.S. courts have the most effective jurisdiction over domestic companies, yet most of these entities are foreign. Whether they will voluntarily comply with U.S. court orders remains uncertain.

The new battleground for AI training data

Another long-term aspect of this case lies in the publishers’ complaint: they explicitly claim that Anna’s Archive is not only providing free e-books to readers but also serving as a major training data source for AI companies like Meta and NVIDIA to build large language models.

This accusation is still one-sided, untested in court, as no trial has been held—defendants have been absent from the start. But the logical chain it exposes is a question the entire AI industry is avoiding: if the training data for models comes from copyrighted works of unclear origin, does training models on such data constitute indirect infringement?

A contrasting point is the previous Spotify-related case, where Anna’s Archive, after winning a judgment in the music industry, proactively deleted the Spotify-scraped data; but the publishers’ books are still accessible on the site. This difference makes it harder for domain intermediaries to dismiss enforcement orders as “self-corrected.”

The bigger issue is timing: the speed of copyright litigation far lags behind AI companies’ consumption of training data. By the time courts confirm infringement and seek damages, the data has already been compressed into model weights, physically unrecoverable.

Regulation always trails behind technology, but it will catch up. What Anna’s Archive faces today is just the first officially convicted link in the entire copyright liability chain. Who will be next in the defendant’s seat remains the real question this judgment leaves behind.