The Training Admission

The authors' filing quotes an internal note from 2019, written by an OpenAI employee: "We trained GPT-3 on pirated stuff! No sharing that!"

OpenAI's brief (Dkt 1983, p.8): "OpenAI used broader mixtures of data from the public Internet, including compilations of books from Library Genesis used in training GPT-3 (Spring 2020) and GPT-3.5 (late 2021), long before ChatGPT was released in November 2022." The brief was made public on 17 September 2026. The note is the authors' exhibit (their statement of facts, ¶344); its writer is not named here. A federal judge, William Alsup, wrote in Bartz v. Anthropic (23 June 2025) that Library Genesis copies "he knew had been pirated"; OpenAI calls that statement dicta, as reported.

In June 2022, five months before ChatGPT came out, the filing says OpenAI's then vice president of research wrote: "Given how much OpenAI is in the news, now is the right time to excise Libgen from our systems and storage."

Under oath this year, Sam Altman was asked: "You agree that piracy is wrong and illegal?" His answer: "I do." Microsoft's Satya Nadella was asked: "It's illegal to download pirated material, right?" "Absolutely. That's correct."

As the authors record them (¶370 and ¶382). The authors also record (¶384) that Nadella "wish[ed] that OpenAI had not gone to a pirated site".

OpenAI's answer: the downloading "was part of a single transformative use". In other words, fair use. It says no other model in the case was trained on the library. Microsoft says it "had no involvement at all in acquiring" the datasets.

OpenAI's brief, its heading: "2. OpenAI's downloading was part of a single transformative use." "No other LLMs at issue in this case were trained on material downloaded from Library Genesis" refers to the seven models in the case, GPT-3 to GPT-4o mini, not every later model. Microsoft (Dkt 1973, p.24): "Microsoft did not download—and indeed had no involvement at all in acquiring—the datasets from LibGen that OpenAI used in connection with the development of its early models." The authors allege Microsoft knew in April 2019.

Anthropic, which makes Claude, the model Token Runner is built with, downloaded from pirate libraries too. In a 2025 ruling, a judge wrote that Anthropic "pirated over seven million copies of books." It settled with the authors for one and a half billion dollars, about three thousand dollars a work, without admitting it broke the law, and it must destroy the files. It says the pirated sets weren't in any model it released. And the authors' filing against OpenAI quotes Anthropic's chief executive, Dario Amodei, in 2020, when he was at OpenAI: "Is it sketchy to call our corpuses 'Books1' and 'Books2' and not say what they are?"

A judge has ordered Meta to name everyone who approved downloading pirate-library datasets holding one publisher's books, or swear it has found no one. Nvidia says contact with a pirate library doesn't mean it took the authors' books. Its case is going ahead.

No court has yet ruled on whether OpenAI's training on those books was fair use. But OpenAI now says, in its own brief, where those books came from.

What we couldn't confirm: Nvidia's motion in its own text (quoted here as TorrentFreak quotes it), and the authors' oppositions of 23 October, reported by a secondary source. The hearing is expected in early 2027, as the Authors Guild expects.