Seattle Times, Newsday Sue OpenAI, Microsoft; 400 Papers Now in Fight

For copyright litigators and IP counsel, the Seattle Times–Newsday complaint is a real-world test of whether ingestion of journalistic works for AI training can survive fair use defenses—and whether courts will order the extraordinary remedy of destroying model weights. With Microsoft in the suit because Copilot runs on OpenAI's technology, the case also targets the deployment layer, not just the model developer. The demand for destruction of datasets and models pushes the law into uncharted territory where copyright remedies collide with software architecture.

The Seattle Times and Newsday have filed a copyright infringement lawsuit against OpenAI and Microsoft, adding two prominent regional publishers to a rapidly widening legal front over generative AI training. The complaint, filed in September 2026, alleges that OpenAI used the publishers' journalism as training data without permission and that its models sometimes reproduce passages from their reporting in response to user queries. Microsoft is named as a defendant because its Copilot assistant is built on OpenAI's technology, effectively extending the alleged infringement into Microsoft's widely distributed commercial products.

“

The New York Times sued OpenAI and Microsoft in December 2023, and other plaintiffs such as Ziff Davis, Merriam-Webster, and Encyclopedia Britannica have filed similar claims.

This suit is not happening in isolation. The two publishers join nearly 400 local newspapers that recently sued OpenAI and Microsoft, asserting that chatbots reduce the need to visit their sites, undercut subscription revenue, and deprive them of the economic value of original reporting. The Seattle Times and Newsday claim similar harm: by providing direct answers drawn from copyrighted articles, AI systems become substitutes for the publishers' own platforms, causing loss of traffic and revenue. In addition to monetary damages, the plaintiffs seek an extraordinary remedy—destruction of any copies of their works held by the companies, along with training datasets and AI models that incorporate them. That demand, if granted, would not merely compensate; it would require technical dismantling or retraining of models, an outcome with massive implications for the entire AI industry.

The case echoes earlier high-profile litigation. The New York Times sued OpenAI and Microsoft in December 2023, and other plaintiffs such as Ziff Davis, Merriam-Webster, and Encyclopedia Britannica have filed similar claims. The fact that both major national publications and a cluster of local publishers are now litigating suggests the news industry has moved beyond negotiation toward legal enforcement. The inclusion of Microsoft is strategically significant: it targets not only the model developer but also the distribution layer, where Copilot has become integrated into Windows, Microsoft 365, and enterprise software. Plaintiffs across multiple suits are testing whether liability can attach to companies that deploy rather than develop models.

From a legal perspective, the core issues are familiar but unresolved. The copying of works for training may be defended under fair use, but courts have yet to rule definitively at scale. The output reproduction claims may be stronger if plaintiffs can show memorization or verbatim outputs, though OpenAI generally imposes guardrails against direct quotation of paywalled articles. The destruction remedy is especially notable: under copyright law, courts can order destruction of infringing copies, but extending that to model weights and datasets raises difficult technical and evidentiary questions. How does one excise a single publisher's works from a model with billions of parameters? Does retraining suffice, or must the model be destroyed? These questions could force courts to confront the nature of model weights as derivative works.