In the latest filing in its copyright lawsuit against OpenAI, The New York Times has put the accusation in terms meant to sting. The newspaper alleges that OpenAI copied millions of its protected articles to build artificial-intelligence technology worth what the suit calls “astronomical” sums, and that it did so knowing the company’s products posed an “existential threat” to the publishers whose work it was taking.
The filing, reported on September 18, sharpens a case that has become the defining test of whether AI companies may train their models on copyrighted material without payment. The Times and other media organizations are seeking billions of dollars in damages, and the central question is whether OpenAI and its co-defendant Microsoft violated copyright law by feeding publisher content into their models.
The Times’ lawyers point to the founding documents and early writings of OpenAI’s leadership to argue that the company’s motives were commercial from the start. They cite Greg Brockman, the co-founder, who around 2017 wrote that he was “deeply driven” by the prospect of “astronomical” wealth, and that he hoped to obtain it by commercializing the technology OpenAI was building.
The phrase matters because it echoes the language the Times uses to describe the value of what was taken. The newspaper argues that OpenAI saw the content it scraped not as protected expression but as raw material, and that the company’s ambition to turn that material into a business is the very thing that made the copying deliberate rather than incidental.
OpenAI has denied the claims and has long maintained that training on publicly available text is fair use, the legal doctrine that permits limited use of copyrighted work without permission. The company has also struck licensing deals with a number of publishers, an effort that critics describe as selective and that supporters point to as evidence of good faith.
Microsoft, which has poured billions into OpenAI and supplies much of its computing power, has argued that the claims against it are overstated and that the company should not be held liable for how its partner trains models. The inclusion of Microsoft has turned the case into a broader fight over the economics of the entire AI industry, since a ruling against the defendants would expose every lab that trained on web text.
The stakes are high enough that the outcome will shape the market for news in the digital age. Publishers argue that their journalism is the source material for the models that now answer readers’ questions directly, and that without compensation the industry that produces that journalism will shrink. AI companies counter that requiring payment for training would entrench a few large players and slow the technology’s development.
The Times has made this argument more aggressively than any other publisher. It filed the suit in late 2023 and has pursued it through motions and discovery while continuing to negotiate over content deals, and its legal team has framed the case as a matter of survival for an industry that has already lost much of its advertising revenue to the internet.
The company’s evidence includes what it says are examples of its articles reproduced nearly verbatim by OpenAI’s models, the kind of output the newspaper argues cannot be explained away as coincidence. Those examples, if they hold up in court, undercut the fair-use defense by suggesting the models retain and reproduce the protected work itself.
The mention of Brockman’s 2017 note is aimed at a jury’s sense of motive. The Times wants to establish that OpenAI’s leaders understood from the beginning that the technology would be valuable precisely because it absorbed the work of others, and that the company pressed ahead in full awareness of the harm publishers might suffer.
OpenAI has said it will fight the case and that its models learn patterns from text the way a person learns from reading, rather than storing and reproducing the material. The company’s lawyers will argue that the outputs the Times cites are anomalies and that the broader value of the models comes from understanding language, not from copying any particular article.
The Times is not alone in court. A cluster of publishers, authors, and music companies have filed similar suits against OpenAI, Microsoft, and other AI developers, each seeking to establish that training on copyrighted work requires a license. The cases are watched closely because, whatever the individual verdicts, together they will set the price of an industry’s raw material.
The case is expected to take years to resolve, and either side’s victory would ripple well beyond these parties. A ruling for the publishers would force the entire industry to renegotiate its relationship with the written word, and a ruling for OpenAI would confirm that the training that built modern AI was lawful all along. Both outcomes are still on the table.
