The New York Times v. OpenAI — A Newspaper Takes Generative AI to Court

On 27 December 2023 The New York Times sued OpenAI and Microsoft for copyright infringement in the US District Court for the Southern District of New York. The complaint alleged that the two companies had copied millions of Times articles without permission to build their large language models, and that ChatGPT and Bing Chat (later Copilot) could recite Times journalism nearly verbatim, closely summarise it and mimic its style — drawing readers away from the Times. It is a case brought by a major US news organisation against a generative-AI developer, and as of October 2026 the court has still not decided whether training on the articles was fair use.

What the complaint alleged

According to the complaint, the Times approached Microsoft and OpenAI in April 2023 to raise intellectual-property concerns and explore a resolution with commercial terms and technological guardrails; the talks produced nothing. It noted that the defendants publicly insisted their conduct was fair use serving a new "transformative" purpose, and answered: "there is nothing 'transformative' about using The Times's content without payment to create products that substitute for The Times and steal audiences away from it."

The complaint also argued that search features built on Bing displayed extensive excerpts of Times content and of Wirecutter's product recommendations, cutting into Wirecutter's affiliate revenue, and that ChatGPT's hallucinations — false information attributed to the Times — damaged its reputation.

OpenAI's response

OpenAI answered on 8 January 2024 with a post titled "OpenAI and journalism", built around four points: it collaborates with news organisations; "Training is fair use, but we provide an opt-out because it's the right thing to do"; "'Regurgitation' is a rare bug that we are working to drive to zero"; and "The New York Times is not telling the full story". OpenAI said its talks with the Times had appeared to be progressing constructively through their last communication on 19 December, and cited the European Union, Japan, Singapore and Israel as places whose laws permit training models on copyrighted content.

Its motion to dismiss on 26 February 2024 went further: "the Times paid someone to hack OpenAI's products," the brief said, adding that it took "tens of thousands of attempts to generate the highly anomalous results that make up Exhibit J". The Times does not accept that account; which side is right is for the case to decide.

The April 2025 ruling and the consolidation

In an opinion dated 4 April 2025, Judge Sidney H. Stein rejected most of the motions to dismiss. OpenAI's argument that claims over 2019–2020 training were barred by the three-year statute of limitations failed, and the contributory infringement claims survived. The misappropriation claims were dismissed, and the DMCA section 1202(b)(1) claim against OpenAI in the Times action was dismissed without prejudice. The same opinion ruled on the parallel suits by the Daily News group and the Center for Investigative Reporting.

That April the copyright suits against OpenAI and Microsoft, including the authors' class actions, were gathered before Judge Stein as In re: OpenAI, Inc. Copyright Infringement Litigation (MDL No. 3143). The Times case now proceeds as one of the "News Plaintiffs" cases. In 2026, after the Supreme Court's decision in Cox v. Sony, the news plaintiffs dropped their material-contribution theory of contributory infringement and their trademark dilution claims, according to the litigation blog ChatGPT Is Eating the World.

Users' chat logs become an issue

The case reached ordinary ChatGPT users too. At the plaintiffs' request, Magistrate Judge Ona Wang ordered OpenAI in 2025 to preserve output data. According to OpenAI, the order covered ChatGPT Free, Plus, Pro and Team and API customers without a zero-data-retention agreement, but not Enterprise or Edu (Enterprise's exclusion was clarified at a hearing on 27 May). OpenAI called the demand an overreach that would keep even deleted conversations, and said its obligations under the order ended on 26 September 2025.

On 7 November 2025 Judge Wang ordered OpenAI to produce a de-identified sample of 20 million ChatGPT conversation logs to the news plaintiffs, and on 5 January 2026 Judge Stein overruled OpenAI's objections and affirmed. His order said that users' privacy interest in conversations they had voluntarily disclosed to OpenAI, and which OpenAI retains in the ordinary course of business, is weaker than the interest in wiretapped phone calls. The case put a new question in front of users: when can a conversation with an AI become evidence?

What each side has at stake

The Times is defending a business model in which reporting costs are recovered through subscriptions and advertising; if AI answers remove the need to visit the article, that recovery weakens. OpenAI argues that AI development depends on training being fair use. In September 2026 the US Justice Department filed a statement in the case that, as Reason reported, said training AI models on copyrighted material is not "in and of itself" infringement and warned that the Times' position could severely hamper technological progress.

Meanwhile, licensing between publishers and AI companies spread. OpenAI pointed early on to partnerships with the Associated Press and Axel Springer, and on 29 May 2025 the Times itself announced its first generative-AI licensing deal — with Amazon, covering display of its content in Amazon products and use for training Amazon's models. Lawsuits and licences are not opposites: both are ways of answering whether, and how much, an AI that uses journalism should pay for it.

In September 2026 both sides filed cross-motions for summary judgment on fair use. Unlike Bartz v. Anthropic, which held training on books to be fair use, this case squarely asks whether the outputs substitute for the articles. The parallel fight over images is told in Getty Images v. Stability AI.