AI copyright lawsuits over pirated books, company by company

Company by company

A court record is an order or the case record; a company’s own word is its filing or paper; an allegation is a claim not yet proved.

Anthropic2021–2022

Court record: Downloaded Books3 (early 2021), more than five million LibGen books (June 2021) and more than two million from the Pirate Library Mirror (July 2022). Bartz, order of June 23, 2025
Court record: $1.5 billion settlement for the 482,460 works on its Works List, approved July 20, 2026; 350 owners opted out, covering 1,802 works. Bartz, Dkts. 680, 619-3
Company’s own word: Argues the opt-out authors’ suits come too late: a three-year limit, it says, bars claims over the 2021 and 2022 downloads. Cambronne, Dkt. 228
Where it stands: First settlement payments expected by Nov. 15, 2026; two appeals pending; hearing on the opt-out suits Dec. 17, 2026.

Meta2022–2024

Court record: First used a shadow library in October 2022 (LibGen) and downloaded Anna’s Archive in early 2024. Fair use granted to Meta on the named authors’ record. Kadrey, Dkt. 598
Company’s own word: Admits downloading LibGen in about October 2022, torrenting Anna’s Archive data, and that some data was uploaded while it torrented (it says de minimis). Sullivan, Dkt. 85; Elsevier, Dkt. 78
Allegation: Publishers: 134.6 TB torrented April–July 2024 and 40.42 TB uploaded. Textbook authors: about 4,500 users got LibGen files from Meta in 2022. Elsevier and Sullivan complaints
Where it stands: Several suits in San Francisco; four set for trial May 24, 2027. Case by case

LibGenAnna’s ArchiveBooks3Full record and sources →

OpenAI2018–2020

Court record: Undisputed, per a discovery order, that an OpenAI employee downloaded pirated books from LibGen in 2018. Authors Guild v. OpenAI, Dkt. 782
Allegation: Two employees torrented about 35 TB, all of LibGen, between September 2019 and January 2020; datasets called LibGen1 and LibGen2 were renamed Books1 and Books2. MDL Dkt. 1987; Dkt. 782
Where it stands: Consolidated in New York before Judge Sidney H. Stein; OpenAI’s summary judgment motion is pending.

Other companies on the record

Suits and papers about other companies’ use of pirate book collections. Each development is graded on our news page.

Nazemian v. NVIDIA, N.D. Cal. 4:24-cv-01454

Court record: Claims over the Pirate Library Mirror, Bibliotik and BitTorrent go forward (May 5, 2026); a ruling on the pleadings, not a finding of copying.
Company’s own word: Its answer admits downloading data from Anna’s Archive’s datasets page and from libgen.rs, without dates (June 9, 2026).
Allegation: Trained on Books3 through The Pile; approved high-speed access to Anna’s Archive in August 2023.

Hachette v. Google, S.D.N.Y. 1:26-cv-05870

Allegation: Publishers and authors allege Gemini was trained on Google Books copies and on web text in C4 that includes pirate sites (July 10, 2026).
Court record: Severed from the authors’ suit against Anthropic (June 8, 2026).

Huckabee v. Bloomberg, S.D.N.Y. 1:23-cv-09152

Company’s own word: The BloombergGPT paper lists Books3 among its training data (Mar. 30, 2023).

Claims severed from Carreyrou v. Anthropic

Allegation: Authors say there is a reasonable basis to believe each used Anna’s Archive (Mar. 2026).
Court record: Claims severed from the suit against Anthropic (June 8, 2026).

No U.S. suit tracked here

Company’s own word: The DeepSeek-VL paper says it used e-books from Anna’s Archive (Mar. 8, 2024).

Lawsuits, company by company