The New York Times first sued OpenAI and Microsoft in December 2023, claiming that millions of its articles were copied without permission to train AI models that now power ChatGPT and Microsoft Copilot. The core of that case has survived multiple attempts to dismiss it. What has changed this week is the shape of the claim against Microsoft specifically.
The Times is now asking the court for permission to file an amended complaint that argues Microsoft did not just provide money or general cloud services. It built a custom supercomputing system, one of the most powerful in the world, designed and built exclusively for OpenAI. Microsoft said as much itself in a public announcement several years ago: the machine was hosted in its Azure cloud and "designed specifically to train" OpenAI's AI models.
This distinction is now legally essential. On March 25, 2026, the US Supreme Court ruled unanimously in Cox Communications v. Sony Music that providing a service while knowing it will be used for infringement is not enough to make you liable. The Court held that contributory liability requires either actively inducing infringement or providing a tool that is purpose-built for infringing activity. Simply knowing and continuing is not sufficient.
The old legal test, called "knowledge plus material contribution," had been the standard in many US courts for decades. The Supreme Court eliminated it. Now plaintiffs must prove intent, which is a higher bar.
The Times has read this ruling and adjusted its argument accordingly. Rather than claiming Microsoft knew and helped, it is now claiming Microsoft actively built the instrument of infringement. A supercomputer custom-designed for one specific company's AI training, the Times argues, is not like an internet provider giving everyone access to the web. It was tailored to one purpose, for one customer, on one mission.
How strong this argument is will depend on what the internal evidence shows. Microsoft CEO Satya Nadella said publicly that the Azure team did the "heavy lifting" in building the computing infrastructure beneath OpenAI's models. The Times claims discovery has turned up new evidence beyond those public statements.
The broader legal picture is moving fast. Just two days before the Times filed this amendment, a coalition representing nearly 400 local and regional US newspapers filed its own lawsuit against Microsoft and OpenAI in the same court. That group accuses the companies of systematically bypassing paywalls and copying articles onto their servers to train AI systems without permission or payment.
OpenAI and Microsoft have consistently argued their use of online content is protected by fair use, the legal principle that allows limited use of copyrighted material without permission when the purpose is sufficiently transformative. Two separate federal judges have recently found AI training to be fair use in other cases, which OpenAI has cited publicly. But those rulings involved different plaintiffs and different facts, and none of them addressed the specific Microsoft infrastructure argument the Times is now making.
For businesses that produce, rely on, or license any kind of written content, this case is worth watching closely. The question at its center is whether the companies that built the infrastructure of modern AI owe the people whose work trained it. That question will not be resolved quickly, but each new filing narrows the range of possible outcomes.