Trump administration backs OpenAI in NYT copyright battle

You know copyright battles are getting wild when the federal government steps in to defend ChatGPT against America's most famous newspaper.

This week, the Trump administration officially intervened in The New York Times' high-stakes copyright lawsuit against OpenAI and Microsoft. In a formal statement of interest, government attorneys threw their weight behind the AI lab, arguing that training large language models on copyrighted articles falls squarely under the fair use doctrine. Restricting AI training, the filing warned, would contradict core copyright principles and actively hinder American prosperity and scientific progress.

The White House goes all-in on fair use

The Times kicked off this battle in December 2023, demanding billions in damages and alleging that OpenAI unlawfully vacuumed up its reporting to train AI systems. But federal attorneys told the court that the publisher's stance goes too far. By trying to narrow fair use to exclude AI training, the administration argued, the Times is undermining the fundamental constitutional goal of copyright law, which is promoting the progress of science and useful arts.

While government lawyers conceded that fair use assessments depend on the specific facts of each dispute, they warned that slapping broad copyright liability on AI training without licenses would be legally incorrect and disastrous for innovation. The filing emphasized that LLMs are already powering breakthroughs across scientific fields, and tying developers' hands would stifle critical research and economic mobility.

Intervening in private litigation via statements of interest has become a favored playbook for the administration, with one official describing the tactic as incredibly successful at cementing policy priorities. It also aligns with the White House's National AI Legislative Framework, which previously argued AI training should count as fair use. It certainly doesn't hurt that Trump harbors longstanding animus toward the Times, against which he is currently waging a separate defamation suit.

A fractured publishing landscape

The outcome of this showdown will ripple far beyond OpenAI and the Times. A growing list of publishers, including the Chicago Tribune, New York Daily News, and the Center for Investigative Reporting, have launched their own copyright challenges against AI labs.

Courts have already sent mixed signals. In a milestone 2025 ruling, a judge decided Anthropic could legally train models on lawfully purchased books, yet remained liable for pirated texts, culminating in a massive $1.5 billion settlement with authors.

Meanwhile, the media industry's unified front has completely dissolved. Dozens of publishers, such as Vox Media, The Associated Press, and Axel Springer, have opted to cash in on licensing partnerships with OpenAI. Even The New York Times broke its own taboo in 2025 by striking a licensing agreement with Amazon to feed editorial articles and recipes into generative AI features.

Why it matters

For AI developers, this filing is a massive breath of fresh air. If the court aligns with the federal government's position, it establishes critical legal cover for scraping public web data to build frontier models. Without the looming threat of billions in statutory damages, teams can keep iterating without needing a licensing deal for every single paragraph their scrapers touch.

For creators and publishers, however, it is a devastating prospect. We think categorizing foundational training as blanket fair use strips content creators of their primary leverage against big tech. If AI models can absorb decades of investigative journalism for free, media outlets are left with almost zero bargaining power to demand fair compensation.

Strategically, the administration is clearly prioritizing AI dominance over legacy intellectual property protections. Framing LLM development as vital to American prosperity signals that Washington views training restrictions as an unacceptable drag on national competitiveness. While this posture accelerates technological development in the short term, it creates a dangerous paradox: if training models for free starves original reporting, the frontier models of tomorrow will eventually run out of quality data to learn from.