News Outlets Challenge OpenAI and Microsoft Fair Use Defense

The New York Times and The Mercury News have filed motions for summary judgment in Manhattan federal court, demanding that a judge rule against OpenAI and Microsoft’s “fair use” defense regarding the training of large language models (LLMs) on copyrighted journalism [1]. The outcome of this case, which centers on whether tech companies can legally ingest news reporting to build AI products, will determine the economic future of digital publishing and the legal boundaries of AI development [1].

The legal battle over training data

The news organizations argue that the defendants’ use of their reporting is not “transformative” but rather a form of “free riding” that threatens the incentives required to produce original journalism [1]. In their filings, the publishers contend that OpenAI and Microsoft have systematically pirated content, including articles hidden behind paywalls, to fuel their AI pipelines [1]. The plaintiffs claim that the cost of human-produced reporting is being undermined by AI tools that can generate 1 million news-style articles for roughly $6,800, a fraction of the cost of maintaining a professional newsroom [1].

Microsoft and OpenAI maintain that their use of public information is protected under the fair use doctrine, which allows for the use of copyrighted material for transformative purposes [1]. Microsoft argues that its Copilot assistant uses internet search capabilities to synthesize information rather than replacing protected expression [1]. OpenAI’s legal team further contends that facts themselves are not subject to copyright, stating that their technology merely makes existing information more accessible to users [1].

Government intervention

The legal landscape shifted on September 1 when the U.S. Department of Justice (DOJ) filed a statement of interest in the case, siding with the tech companies [2]. Associate Attorney General Stanley Woodward argued that training an AI model on copyrighted text is “extraordinarily transformative” and does not constitute infringement because the training process does not reveal the underlying content to the public [2]. While this filing is not binding on Judge Sidney Stein, it represents the first time the federal government has intervened in the ongoing wave of AI copyright litigation [2].

The news organizations have also accused OpenAI of destroying evidence related to its ability to locate specific news stories within its training data [1]. While the summary judgment motions currently focus on Microsoft’s activities, the publishers are preparing to unseal documents they claim contain “scorchingly hot” admissions from Microsoft executives that could undermine the defendants' arguments [1].

What to watch

• Court rulings: Judge Sidney Stein’s decision on the pending motion for sanctions against OpenAI regarding the alleged destruction of evidence [1].
• Unsealing of evidence: The potential release of redacted internal communications from Microsoft and OpenAI that the plaintiffs claim will serve as a “smoking bazooka” against the fair use defense [1].
• Broader industry impact: Whether the DOJ’s stance on text-based AI training influences pending copyright cases involving music and creative arts, where plaintiffs are increasingly shifting their focus toward AI-generated outputs rather than just training inputs [2].

The case now rests on whether the court views the ingestion of human-authored content as a transformative technological advancement or a violation of the intellectual property rights that sustain the news industry. With the DOJ signaling support for the tech sector’s training methods, the plaintiffs face a high hurdle in proving that the front-end of the AI pipeline constitutes copyright infringement.