OpenAI, Microsoft, and a coalition of news publishers including The New York Times filed dueling summary judgment motions in Manhattan, asking a federal judge to decide whether training large language models on copyrighted journalism is fair use before the dispute reaches a jury.
The filings, submitted around September 4 and widely reported into the new week, crystallize years of discovery into a single legal question with industry-wide stakes. Publishers argue the companies copied millions of articles at scale to build commercial products that substitute for original reporting. OpenAI and Microsoft counter that model training is transformative, that verbatim regurgitation is limited, and that a ruling against them would chill U.S. AI competitiveness.
The U.S. Department of Justice has already filed a statement of interest supporting the fair-use framing, citing innovation and national-security concerns if American labs are forced into fragmented licensing regimes while foreign competitors train freely. That intervention raises the political temperature around what would otherwise be a private copyright docket.
Microsoft’s exposure is structural, not symbolic. The company has poured more than $10 billion into OpenAI and wires its models through GitHub, Office, and Bing. A finding of liability would not only threaten damages and injunction risk; it would reopen the economics of every enterprise AI bundle sold on the promise of unlimited knowledge drawn from the open web.
Judge Sidney H. Stein’s forthcoming decision will not end the appellate path, but it can shrink the case to a few triable facts—or extinguish it. Until then, every publisher negotiating AI licenses and every lab scraping the public internet is trading under a cloud that Congress has so far refused to clear with legislation.
Image source: i.ibb.co