Definition
AI copyright encompasses legal disputes over using copyrighted content to train large language models, the fair-use defense, publisher economic harm from AI products, and licensing frameworks for training data.
Authors Guild v. OpenAI/Microsoft (September 2026 Summary Judgment)
- 2026-09-17: Partial summary judgment motion filed with 162-page statement of facts (2026-09-27-authors-guild-microsoft-openai-piracy-briefs)
- Allegations: intentional library-genesis piracy; Alec Radford downloaded ~117,500 books (2018); Microsoft knew as early as April 2019 Gates presentation
- “Project Clear” (June 2022): internal effort to excise LibGen from systems
- Jack Clark: GPT would “make people unemployed”; Tarun Gogineni aimed to autocomplete G.R.R. Martin series
- 194 books in motion; hearing expected early 2027
Plaintiff filings and press release — defendants dispute liability and maintain fair-use defense.
NYT v. OpenAI/Microsoft (September 2026 Escalation)
Unsealed summary judgment filings revealed internal admissions undermining defendants’ fair-use defense:
- Microsoft Brent Hecht: Called AI scraping “the largest theft of labor in human history” and “an astonishing theft of unprecedented proportions”; said plan made “a complete mockery of ‘fair use’”
- OpenAI leadership: Head of ChatGPT wrote publishers face “existential threat” from “largely substitutive” AI products
- Scale: 91,692+ publisher works in mid-training datasets; 2.06M nytimes.com documents in Common Crawl; Project Mango dataset with 160,903 unique publisher works
- Economic harm: Copilot reduced NYT click-through rates 87–93% below traditional Bing
- Allegations: Paywall circumvention, copyright notice stripping, Bing dataset repurposing without consent
Allegations are from plaintiff court filings; defendants dispute liability and maintain fair-use defense. Trump administration filed brief supporting OpenAI's unlicensed training use.
Developer Implications
- Training data licensing uncertainty affects all teams building on LLMs
- RAG architectures and publisher partnership models as risk mitigation
- Outcome will set precedent for global AI training practices
Related
- fair-use
- ai-training-data
- new-york-times
- openai
- microsoft
- project-mango
- creator-rights-vs-ai
- authors-guild
- library-genesis
- training-data-copyright
- openai-copyright-mdl