Trump Administration Backs OpenAI's Fair Use Defense in Publishers' Copyright Cases

The Trump administration has formally backed OpenAI’s position that training AI models on publishers’ copyrighted content qualifies as fair use, filing a statement of interest in the Southern District of New York in cases brought by The New York Times and Ziff Davis. For AI builders, this intervention signals that the federal government views broad training data access as a national competitiveness issue, not just a copyright dispute.

The government’s filing supports OpenAI’s claim that its use of publishers’ content to train AI models constitutes fair use. The statement argues that a ruling against OpenAI would hinder AI development, limit competition among AI firms, and harm U.S. competitiveness against foreign adversaries. “Rules of law that make it significantly more difficult to develop a robust AI industry in the United States … threaten national security,” the filing states. The government also warned that licensing requirements could create “barriers that function primarily as large subsidies for old mainstream media companies.”

The New York Times sued OpenAI in 2023, alleging the company used millions of its articles to train ChatGPT without permission. Ziff Davis (CNET’s parent) filed a similar suit in 2025. The administration’s brief does not decide the cases, but it adds a powerful voice to the fair use side. OpenAI rival Anthropic previously settled similar piracy claims with authors for $1.5 billion, though a judge in that context ruled the company’s use of copyrighted material was fair use.

Why this matters for AI builders

If the court adopts the government’s reasoning, it could reduce legal uncertainty for AI teams training on publicly available web content. The filing explicitly argues against “licensing entry barriers” that would favor large media companies and reduce competition. That’s directly relevant to founders and developers building LLM-based products who rely on diverse training data sources.

A broad fair use ruling would mean AI companies likely do not need to negotiate licenses with every publisher whose content appears in training sets. It would also lower the cost of assembling training data, since licensing fees can be significant. The government even cited independent authors using AI models to generate images that might otherwise require a license, framing training data access as a pro-competition measure.

But publishers argue the opposite. A New York Times spokesperson told CNET that the administration is “siding with a handful of trillion-dollar AI companies” and that AI companies should “pay fairly for the content that makes their products possible.” The outcome will shape whether AI builders need to budget for data licensing or can continue training on the open web.

Practical implications and risks

For now, AI builders should not assume the legal landscape is settled. The DOJ’s statement of interest is persuasive but not binding. The cases are ongoing, and other lawsuits (including from authors like George R.R. Martin and the Tumbler Ridge shooting tragedy) present separate legal theories. The Anthropic settlement shows some AI companies are paying to avoid litigation risk even when fair use is a plausible defense.

If you are training a model today, you can still use publicly available content, but you should monitor these cases. A future ruling against OpenAI could retroactively change the legal risk for past training. Consider documenting your data sourcing and, if possible, using opt-out mechanisms for publishers who request removal.

Caveats to keep in mind

The filing is a statement of interest, not a court ruling. The fair use defense in AI training is not yet settled law. The government’s position may shift with a new administration. And the cases involve specific plaintiffs; other copyright holders may still sue. The safest path for builders is to stay informed and prepare for either outcome.