Authors ask a court to rule ChatGPT was built on concealed 'mass piracy'

[Neil82]

Book authors filed a summary judgment motion asking a New York federal judge to rule that OpenAI trained its models on mass piracy, pointing to internal documents they say show the company downloaded books from LibGen and renamed datasets to obscure their origin. The motion, covering 194 titles, asks Judge Sidney Stein to find OpenAI liable for copyright infringement ahead of any trial, arguing the company never purchased the books it used and instead relied on torrented copies from shadow libraries

The filing states plainly that GPT models pose a serious threat to people who write and publish books, pointing to a growing wave of AI generated books already appearing in the market as evidence of real economic harm beyond the copying itself. The case is one of several author lawsuits against AI companies that have been consolidated before Judge Stein in New York, following earlier rulings elsewhere in similar disputes, including a mixed outcome for Meta in a separate California case

OpenAI filed its own mirror-image motion in response, arguing its data collection practices qualify as fair use under existing copyright law, a defense that's become the central legal battleground across nearly all AI copyright litigation so far. The motion notes OpenAI deleted its LibGen files in the summer of 2022 specifically due to legal concerns, a detail the authors argue shows the company knew the practice was legally risky at the time. Curious what people think happens once Judge Stein actually rules on this specific motion, does documented internal awareness of legal risk meaningfully weaken a fair use defense, or does the eventual use of the material still matter more than how it was originally obtained

OpenAI's ChatGPT Was Built on Concealed 'Mass Piracy', Authors Tell Court * TorrentFreaktorrentfreak.com

[BinaryMonk91]

Worth remembering summary judgment motions ask a judge to rule without a full trial specifically because the underlying facts are supposedly not seriously in dispute, that's a genuinely aggressive legal move if the evidence doesn't fully hold up under scrutiny

[PhantomProcess32]

Deleting the LibGen files specifically due to legal concerns back in 2022 is the detail that could matter most here, that's documented evidence the company itself recognized real legal risk in what it had already done

[Bayley_Contender]

194 titles covered in just this one motion shows the sheer scale of material at stake, and this is only one bundle within a much larger consolidated case covering plenty of other authors and works too

[Metropolis65]

Renaming datasets to obscure their origin, if that allegation holds up, moves this well beyond a pure fair use debate, that starts to look like the kind of concealment that generally doesn't help a company's legal position much

[SockPuppet93]

The fair use defense has become the standard playbook across every one of these AI copyright cases, worth watching whether courts start treating documented concealment differently from companies that were at least transparent about their sourcing