AI training copyright litigation (DOJ / OpenAI)
Where it stands: AI training copyright is a widening pile of lawsuits: the Justice Department has urged a judge to side with OpenAI, while Seattle Times and Newsday sued OpenAI and Microsoft and USA Today Co. sued OpenAI, seeking more than $250 million. Unredacted filings quote a Microsoft executive calling OpenAI's data practices 'the largest theft of labor in human history,' and Wikimedia has asked to weigh in with a middle ground: protect training but not regurgitation. Watch whether courts adopt that split and more platform operators file.
Latest
- 2026-10-30(The Verge — theverge.com/policy + unite.ai — unite.ai): the platform-operator side enters the training-copyright lawsuits — the Wikimedia Foundation asked a federal court for leave to file an amicus brief (announced October 30, 2026), urging a middle ground: legal protection for LLMs training on copyrighted work, but not for chatbots regurgitating that work to users; an overly black-and-white ruling "could either deprive Wikipedians of useful AI-based tools or endanger the open web." The operator previously vocal about AI agents accessing its platforms is now a legal participant in the copyright pile-up — the first amicus mover in our entry whose posture is pro-training, anti-regurgitation. Watch: whether the leave-to-file motion is granted, whether the training/regurgitation split travels into any ruling, and other platform-operator filings.
AI image - 2026-10-08 the newspaper front gains a national chain — USA Today Co. + owned local papers sued OpenAI (filings Thursday; Verge 1:58 PM ET, per Reuters): "hundreds of thousands" of articles allegedly copied for training, damages sought more than $250M — the first specific damages figure in our record. The pile-of-plaintiffs shape continues (Seattle Times/Newsday, Sep 6); legal backdrop unchanged (Mehta antitrust dismissal Oct 1, Third Circuit generative carve-out). Watch: the damages theory's arithmetic, other chains filing, OpenAI's answer.
- 2026-10-01 (TechTimes report; single outlet — read the opinion before weight): training fair use loses its first appellate test — reported: the Third Circuit ruled Sep 29 against a non-generative AI's fair-use defense, with an explicit carve-out for generative AI training, shape-consistent with the Thomson Reuters v. Ross line. The carve-out cuts both ways: a loss for training defendants that hands their successors first appellate cover simultaneously. Watch: the opinion text (not yet in our feed), wire pickup confirming the case, which front cites it first.
September 2026 · 6 updates
- 2026-09-25 (Suno strand): Sony and UMG sued Suno again (Verge, pub 11:51 AM ET) — the new suit claims v6 still infringes because it is trained on user outputs from previous models, which were themselves trained on unlicensed music. Directly tests the Sep 9 "licensing turn": a training set built from a dirty model's outputs is the labels' counterexample to the licensed-model defense. Watch: whether the suit details v6's training pipeline, and whether Suno answers with a genuinely clean base.
- 2026-09-17 the filings turn against both — newly unredacted court filings (TechCrunch, ~4 PM ET) show a Microsoft exec privately called OpenAI's data practices "the largest theft of labor in human history" while both companies scraped; "doom loop" admissions (AI news products starving the web that feeds them) in executives' own words. Filing-grade admission-shape material from inside the co-defendant.
- 2026-09-10 the turn's second day, from the label side: UMG is launching its own AI music platform on a multiyear licensing deal with ElevenLabs (Verge, 11:38 AM ET) — licensed catalog remixes/mashups, artist opt-in as the consent mechanism. The biggest litigant is now an operator. Watch whether the sued labs' licensed-data equivalents (if any) adopt the same artist-consent terms.
- 2026-09-09 the licensing turn — Suno v6 (Verge 5:42 PM ET) is the first music model "made with support from the record industry," trained from the ground up on licensed content, explicitly not the prior training set. The sued company shipping the counter-model is the shape the thread's resolution had to take if the labels' framing wins. Watch for: label responses (settlement vs continuing suit), whether v6 output quality is judged acceptable, whether other sued companies (OpenAI in the DOJ/publisher fronts) build licensed-data equivalents.
- 2026-09-08 the executive branch itself is reportedly split on whether AI should train on Americans' copyrighted work (Heartlander News) — single low-tier source, verify before weight; if it holds, it complicates the DOJ brief's posture in the same case.
- 2026-09-06 Seattle Times and Newsday sued OpenAI and Microsoft for copyright infringement — a new front: daily newspapers. The pile-up of fronts (DOJ brief → publishers) is becoming the story itself.
Background
- The thread: DOJ urged a federal judge to side with OpenAI on copyright protection for AI training — government weigh-in on the central legal question of the training-data fights.
- Watch for: the judge's response, which case this lands in, whether other branches pile in.