NYT hits Microsoft for building copyright-infringing supercomputer to aid OpenAI
In an updated lawsuit, The New York Times alleges Microsoft built a supercomputer specifically designed to help OpenAI train AI models on copyrighted works without permission, disproportionately weighting Times articles.
The New York Times has filed an amended complaint against OpenAI and Microsoft, focusing on a supercomputer that Microsoft constructed. The newspaper argues that the machine was not a generic cloud computing system but was tailor-made to help OpenAI infringe copyrights by training AI models on protected works without authorization.
According to the NYT, Microsoft designed this “unusually complex” supercomputer with the explicit goal of using nearly the entire internet—curated to feature Times works disproportionately—to create the most capable large language model (LLM) in history. Both companies hoped to train models on high-quality journalism to enable confident mimicry of that writing style. The NYT alleges that Microsoft not only helped select the works that were infringed but also provided the means to seize copyrighted works.
During discovery, outputs were shared that showed near-verbatim excerpts of NYT articles. In some instances, users tried to bypass paywalls by asking for the “next paragraph,” and the model complied. In other cases, the model regurgitated several paragraphs without such manipulation. The NYT claims this substitution causes market harm by positioning ChatGPT as a replacement for a NYT subscription.
Additionally, the NYT alleges that Microsoft’s deployment of Times-trained LLMs across its product line has boosted its market capitalization by a trillion dollars in the past year. The complaint includes side-by-side comparisons and screenshots of allegedly infringing outputs as evidence.
