The lawsuit count against OpenAI just hit a number that would make any insurance underwriter quit: nearly 400 local newspapers plus two more major metros, all claiming the same thing.
The Summary
- The Seattle Times and Newsday sued OpenAI and Microsoft, alleging unauthorized use of journalism as training data and reproduction of reporting in AI responses
- They join nearly 400 local newspapers already suing the same defendants, plus prior suits from The New York Times, Ziff Davis, Merriam-Webster, and Encyclopedia Britannica
- The timing pushes back on Trump administration claims that AI benefits local news
- Microsoft gets named because Copilot runs on OpenAI tech, making this about the entire consumer AI stack
The Signal
The core claim is identical across hundreds of plaintiffs: OpenAI scraped journalism without permission, used it to train models, and now reproduces that reporting when users ask questions. What makes this wave different is the sheer volume. Nearly 400 local newspapers filing suit represents a coordination level that suggests organized legal strategy, not isolated grievances.
The defendant list now includes Microsoft explicitly, which matters because Copilot is the consumer face of GPT for millions of Windows and Edge users. If publishers win, the liability spreads beyond OpenAI to every company that builds on top of their models. That's the real legal surface area expanding here.
"Nearly 400 local newspapers filing suit represents a coordination level that suggests organized legal strategy, not isolated grievances."
The timing is sharp. This lawsuit directly contradicts recent Trump administration talking points that AI companies help local news. The administration's argument: AI drives traffic, surfaces reporting, expands reach. The publishers' argument: AI reproduces our work verbatim, replaces the click, and trains on our IP without paying. Those are incompatible views of the same technology, and discovery in these cases will produce data on which one is accurate.
What the plaintiffs want is compensation for past use and licensing terms for future training. What OpenAI wants is fair use protection, the legal doctrine that lets you quote, critique, and build on copyrighted work without permission. The gap between those positions is not a negotiation. It's binary. Either training on published text is fair use or it isn't. Either reproducing passages in responses is infringement or it isn't. There's no middle ground that makes both sides right.
Key precedents being tested:
- Can model training on public text qualify as transformative use under fair use?
- Does reproducing trained material in outputs cross into direct infringement?
- Are AI companies liable for downstream products (Copilot) built on their models?
The list of publishing plaintiffs now includes The New York Times, Ziff Davis, Merriam-Webster, and Encyclopedia Britannica. That's daily news, tech media, reference dictionaries, and encyclopedias. The breadth suggests this isn't about one type of content or business model. It's about whether OpenAI built a foundation model on borrowed IP and whether that borrowing was legal.
The Implication
If you're building on OpenAI models or planning to, watch discovery in these cases. The legal question isn't just "did OpenAI infringe," it's "does everyone who uses their API inherit that liability." Licensing deals will accelerate if publishers start winning. If OpenAI wins on fair use, expect model training to continue at scale with no payments. If they lose, expect a new licensing layer between foundation models and the content they were trained on. That changes unit economics for everyone in the agent economy who assumed training data was free.
For publishers, coordination is the strategy. One lawsuit is a test case. Four hundred is a pattern that forces settlement or sets binding precedent. Either outcome reshapes the next five years of AI development.