OpenAI won a rare boost from the U.S. Justice Department when the agency filed a 20-page amicus brief backing the company’s fair-use defense in the New York Times copyright lawsuit. The filing, lodged in the Southern District of New York, signals that the federal government sees AI training practices as a matter of national interest, not just a private legal dispute.

Why the Government Is Getting Involved

The brief frames the case in geopolitical terms. It argues that the United States has a “strong interest” in keeping its AI industry competitive, linking AI leadership to both economic prosperity and national security. By supporting Open AI, the administration suggests that a narrow reading of copyright law could slow the research and development cycle that keeps American firms ahead of overseas rivals. The position echoes earlier executive orders that call for a U.S.-first approach to AI standards and governance.

At stake is whether feeding massive, copyrighted corpora into a large language model (LLM) counts as “fair use.” Fair use protects uses that are “transformative” – that is, uses that add something new, with a different purpose, rather than simply reproducing the original work. Open AI’s defense, echoed by the government brief, holds that LLMs learn patterns and generate novel text, a process more akin to a human reader absorbing information than a copy-and-paste operation.

The brief warns that “constraining LLM development under a misunderstanding of fair-use doctrine” would damage American economic mobility. That language mirrors a comment from a separate case in which a judge likened LLM training to a reader learning from books to create new ideas, rather than a tool designed to replace the original creator.

How Prior Cases Shape the Debate

The brief does not carry the force of a court ruling, but it points to recent litigation that helps define the line between permissible training and illegal data acquisition. In a case involving Anthropic, the company faced a $1.5 billion settlement not because its model learned from copyrighted works, but because it sourced that material from “illegal shadow libraries” – essentially pirated databases. The settlement underscores that the method of obtaining data can trigger massive liability even if the training process itself might be deemed transformative.

Judge William Alsup’s earlier observations provide additional context. He emphasized that training an LLM is more like a human learning process than a direct copy, suggesting the courts may be open to a broader fair-use interpretation. Yet the Anthropic outcome shows that courts draw a hard line at illegal procurement, regardless of the downstream use.

Who Gains and Who Risks

AI developers stand to benefit from a legal environment that treats large-scale data ingestion as fair use. A favorable ruling could lower the cost of building next-generation models such as Claude, Gemini, or a future GPT-5, because firms would no longer need to negotiate licenses for every piece of text they ingest. That could accelerate innovation and keep U.S. firms at the forefront of the global AI race.

Content creators and publishers remain the most vocal opponents. They argue that unrestricted scraping erodes the value of their work and deprives them of revenue. The New York Times lawsuit reflects that concern: the newspaper claims Open AI’s use of its articles without permission violates its copyrights. If courts side with the newspaper, AI labs could face injunctions, damages, or the need to retroactively license billions of words—a financial and logistical burden that could choke smaller players.

The government has a stake in balancing these forces. While the brief champions AI growth, it also implicitly acknowledges the need for a clear legal framework. Overly permissive rulings could provoke backlash from the creative sector, potentially prompting new legislation that might be more restrictive than the current dispute.

The Counter-Argument: Protecting Creative Rights

Critics of the government’s stance point out that fair-use doctrine was never intended to blanket an entire industry’s data practices. They warn that treating all large-scale text ingestion as transformative could set a precedent that undermines the copyright system’s incentive structure. Moreover, the Anthropic settlement demonstrates that even if the training process is permissible, the means of acquiring data can still be illegal. This duality means that AI developers cannot simply rely on a fair-use defense; they must also ensure their data pipelines are clean.

What to Watch Next

  • Court rulings: The Southern District of New York will eventually issue a decision on the New York Times claim. That judgment will likely become a reference point for future AI-related copyright cases.
  • Legislative moves: Lawmakers may respond to the court’s direction by drafting statutes that clarify permissible data-use practices for AI, potentially tightening or loosening the current gray area.
  • Industry response: AI firms may adjust their data-collection strategies, either by seeking broader licensing agreements or by building internal datasets that avoid copyrighted material altogether.

Bottom Line

The Justice Department’s amicus brief turns the Open AI copyright fight into a proxy battle over America’s AI future. By framing LLM training as a fair-use activity essential to national competitiveness, the government signals that restrictive copyright interpretations could be a strategic liability. Yet the ongoing tension between creators’ rights and developers’ needs means the courts, and possibly Congress, will have to draw a line that balances innovation with the protection of creative work.