- Major Publishers Challenge AI Training Practices in Landmark Copyright Suit Against Meta (Holland & Knight, 2026-05-05) — Five major publishing houses — Elsevier, Cengage, Hachette Book Group, Macmillan Publishers, and McGraw Hill — plus author Scott Turow filed a putative class action against Meta and Mark Zuckerberg in the SDNY on May 5. The case focuses on two fair use issues not present in author-only suits: unlawful sourcing of training data AND demonstrable market harm (Meta’s Llama allegedly produces full-length scientific papers, replacement chapters, and study guides that substitute for the plaintiffs’ works). This is the first case brought by institutional publishers with robust market data and established licensing programmes — plaintiff profiles that make the market-harm factor materially stronger than in previous suits.
- AI Lawsuits in 2026: Settlements, Licensing Deals, Litigation (AI Business) — Landscape survey of active cases post-Bartz. The trajectory: music publishers’ $3B piracy suit (filed January 29) amends in light of the Bartz settlement; Disney/OpenAI licensing deal ($1B investment + Sora access to Disney characters) signals the parallel licensing market developing alongside litigation. Two strategies are now running simultaneously: sue for damages, or license for investment. Both are real markets.
Litigation Front — Copyright Shifts to Outputs #
- AI Trends for 2026 — Copyright Litigation Shifts from Training Data to AI Outputs (Morrison Foerster) — Morrison Foerster’s 2026 prediction: the training data litigation wave (Bartz, Meta publishers) is peaking; the next wave is AI output liability — the “substitutive summary” doctrine from Judge McMahon’s ruling (already in this journal) will extend to RAG products, AI search, and summarisation tools. The liability surface is expanding even as training-data doctrine clarifies.
- Thomson Reuters v. ROSS — June 11 Oral Argument (BakerHostetler) — The Third Circuit oral argument is set for June 11, 2026 — the first appellate argument testing AI training fair use directly. Both parties filed supplemental briefs on ASTM v. UpCodes (a different Third Circuit case on fair use of legally-incorporated standards) with diametrically opposed readings. The court requested those briefs, signalling active deliberation on how UpCodes affects AI training analysis. Oral argument June 11 implies a decision likely Q3–Q4 2026.
Cross-links #
- Open vs Closed AI Ecosystems The Meta publisher case targets Llama specifically — open-weight model producers now face institutional publishers with established licensing infrastructure as plaintiffs, not just individual authors. The IP exposure asymmetry between open and closed labs is getting larger.
- AI Impact on Society The Disney/OpenAI licensing deal ($1B investment + Sora character access) represents the parallel market: rights holders can choose litigation or commercial partnership. The two paths are not mutually exclusive — different rights holders will choose differently.
- Emerging pattern: The litigation landscape is bifurcating by plaintiff type: individual authors (Bartz, music publishers) → piracy/training-data claims; institutional publishers (Meta case, potentially others) → market-harm + training-data claims. The institutional-publisher cases add a materially stronger market-harm argument that individual author suits lack.
- Quality signal: The Morrison Foerster output-liability prediction (February 2026) is now the leading indicator to watch. If the Thomson Reuters ROSS appeal goes for the plaintiff in Q3, output liability cases will accelerate simultaneously. A two-front opening — training and output — would reshape the entire industry’s legal posture.
- Keyword suggestion:
"market harm" AI output substitution copyright 2026 — the substitutive-summary angle (market harm from outputs replacing originals) is now the active frontier; the training-data question is settling.