The New York Times Company filed a lawsuit against OpenAI and Microsoft on December 27, 2023, in the United States District Court for the Southern District of New York. The complaint alleges that the defendants used millions of copyrighted news articles without permission to train their large language models and that the resulting systems compete with the newspaper's own content. The case is one of the most closely watched legal battles over the intersection of generative AI and copyright law, with implications for the entire artificial intelligence industry.
The lawsuit centers on claims of direct and vicarious copyright infringement, as well as violations of the Digital Millennium Copyright Act. The Times seeks statutory damages, injunctive relief, and the destruction of any models and training datasets that incorporated its works. OpenAI and Microsoft have responded with motions to dismiss, arguing that their use of the articles constitutes fair use under U.S. copyright law, a defense that has yet to be fully resolved in court as of early 2025.
Background and Parties
The New York Times, founded in 1851, is one of the largest and most influential newspapers in the United States, with a digital subscription base exceeding 10 million by 2023. Its business model relies heavily on exclusive reporting and paywalled content. OpenAI, established in 2015, is a leading AI research organization that developed the GPT series of models, including GPT-3 and GPT-4, which power products like ChatGPT. Microsoft, a major investor in OpenAI with a reported $13 billion commitment, provides cloud infrastructure through Azure and integrates OpenAI's models into its Bing search engine and Office productivity suite.
The complaint identifies over 100 specific examples where ChatGPT and Bing Chat allegedly reproduced Times articles nearly verbatim, often behind paywalls, without attribution or licensing. The Times argues that this undermines its ability to monetize its journalism, as users can obtain the same information from the AI systems without subscribing.
Legal Claims and Allegations
The primary claim is copyright infringement, asserting that the defendants copied, reproduced, and distributed the Times' articles during the training of their models. The complaint details how the training process involves ingesting massive text corpora scraped from the internet, which included millions of Times articles. The Times further alleges that the models can generate outputs that are substantially similar to its articles, sometimes including exact sentences or paragraphs.
The lawsuit also alleges vicarious infringement, claiming that OpenAI and Microsoft profited directly from the infringing acts of their users, who prompted the models to reproduce copyrighted content. Additionally, the Times asserts a claim under the Digital Millennium Copyright Act, arguing that the defendants removed or altered copyright management information, such as bylines and copyright notices, during the training process.
Defense Arguments
OpenAI and Microsoft filed a motion to dismiss in February 2024, arguing that the lawsuit is based on speculative claims and that the use of copyrighted works for training is a transformative fair use. They contend that the models do not replace the market for the original articles but rather serve as a new tool for information synthesis. The defendants also point to the fact that the Times articles were publicly accessible online, and that the training process involves statistical patterns rather than direct copying.
In their filings, OpenAI has emphasized its efforts to provide attribution and to allow publishers to opt out of training, though the Times argues these measures are insufficient and came only after the infringement occurred. The defendants have also noted that the Times itself uses AI tools for various purposes, suggesting a degree of acceptance of the technology.
Industry Context and Related Cases
The lawsuit is part of a broader wave of copyright disputes involving AI companies. In 2023, authors including Sarah Silverman and George R.R. Martin filed class-action suits against OpenAI and other companies, while visual artists sued Stability AI and Midjourney. Getty Images filed a separate suit against Stability AI in the UK and the US. These cases share common questions about whether training on copyrighted data constitutes fair use, and whether AI-generated outputs can infringe on existing works.
The outcome of the NYT case could set precedent for how courts interpret fair use in the context of machine learning. Legal scholars have noted that the transformative use doctrine, established in cases like Authors Guild v. Google, may be central to the defense. However, the commercial nature of OpenAI's products and the potential market harm to the Times distinguish this case from earlier precedents.
Technical Aspects of Training Data
The training of large language models involves feeding billions of words from diverse sources, including books, websites, and news articles. OpenAI's GPT-3, released in 2020, was trained on a dataset called Common Crawl, which contains petabytes of web pages. The Times alleges that its articles were included in this and subsequent datasets without authorization. The technical process involves converting text into numerical representations using techniques like transformers and multi-head attention, which allow the model to learn patterns and generate coherent responses.
OpenAI has acknowledged that its training data includes publicly available web content but has not disclosed the full composition of its datasets. The company has implemented filters to reduce the likelihood of reproducing copyrighted text verbatim, but the Times' complaint provides examples where these filters failed. The case has prompted discussions about the need for more transparent data sourcing and the potential for licensing agreements between AI companies and content creators.
Procedural History and Status
After the initial filing, the court consolidated the case with other related copyright lawsuits against OpenAI for pretrial purposes, though the Times' case remains distinct. In March 2024, the judge denied the defendants' motion to dismiss in part, allowing the core copyright claims to proceed while dismissing some secondary claims. As of early 2025, the parties are engaged in discovery, with the Times seeking access to OpenAI's training data and internal communications.
A key development occurred in November 2024 when OpenAI filed a counterclaim against the Times, alleging that the newspaper had hired a contractor to manipulate ChatGPT into reproducing copyrighted material, violating the platform's terms of service. The Times has denied these allegations, calling them a distraction from the central issues. The case is scheduled for trial in late 2025, though settlement remains possible given the high stakes for both parties.
Implications for Journalism and AI
The lawsuit has significant implications for the news industry, which has struggled to adapt to the digital economy. Many publishers have entered into licensing agreements with AI companies, such as the deals between OpenAI and Axel Springer, the Associated Press, and News Corp. However, the Times has chosen litigation, arguing that licensing alone does not address the fundamental harm of unauthorized use. The case could determine whether AI companies must pay for the data that underpins their products, potentially reshaping the economics of both journalism and AI development.
For the AI industry, an adverse ruling could force companies to overhaul their training practices, potentially limiting the scale of datasets and increasing costs. Conversely, a ruling in favor of OpenAI could establish broad fair use protections, encouraging further innovation but raising concerns among content creators. The case is closely watched by researchers at institutions like MIT CSAIL and Stanford AI Lab, who study the legal and ethical dimensions of AI.
Public and Scholarly Reactions
Public opinion on the lawsuit is divided. Journalists and media advocates generally support the Times, viewing it as a defense of intellectual property and the sustainability of quality journalism. Tech enthusiasts and some AI researchers argue that restrictive copyright rulings could hinder progress, pointing to the long history of using existing works to create new knowledge. Legal experts have noted that the case may ultimately reach the Supreme Court, given the novelty and importance of the questions involved.
Scholars have also debated the concept of fair use in the context of AI, with some proposing new frameworks that balance innovation with creator rights. The case has spurred academic conferences and papers, including analyses of how neural networks memorize training data and the conditions under which reproduction occurs. This research could inform judicial decisions and future legislation.
Potential Outcomes and Future Directions
Several potential outcomes are possible. The court could rule that training on copyrighted works is fair use, which would likely lead to more licensing negotiations but not mandatory payments. Alternatively, the court could find infringement, requiring OpenAI to pay damages and possibly destroy its models, which would be unprecedented and disruptive. A middle ground might involve a ruling that allows training but requires stronger safeguards against verbatim reproduction, or that mandates a licensing regime.
Beyond the courtroom, the case has prompted legislative proposals, such as the AI Foundation Model Transparency Act, which would require companies to disclose training data sources. The European Union's AI Act, passed in 2024, includes provisions on copyright compliance for training data. These regulatory efforts may provide a framework that reduces the need for case-by-case litigation. The NYT v. OpenAI lawsuit thus stands as a landmark case that will influence how society balances the benefits of AI with the rights of creators for years to come.