No, Trump Can Not Decide That OpenAI’s Training Is Fair Use

The Trump administration has entered The New York Times copyright lawsuit against OpenAI—but not in the way some headlines might lead readers to believe.

The administration did not issue a ruling. It did not rewrite copyright law. It did not order the court to protect OpenAI. The Department of Justice filed a statement of interest arguing that courts should interpret fair use in a way that does not unnecessarily impede American artificial-intelligence development.

That is advocacy, not authority.

Which branch actually controls the outcome?

All three branches have different roles here.

Congress, the legislative branch, writes federal copyright law. It created copyright owners’ exclusive rights in 17 U.S.C. § 106 and the fair-use limitation in 17 U.S.C. § 107.

The executive branch—here, the Trump administration acting through the Department of Justice—enforces federal law and may tell a federal court how the United States believes that law should be interpreted. Congress expressly authorizes the attorney general to send Justice Department lawyers into pending cases to protect the interests of the United States under 28 U.S.C. § 517.

The judicial branch decides the lawsuit. Because the case is currently in the U.S. District Court for the Southern District of New York, the district judge decides the issues presented at this stage. Any appeal would ordinarily go to the U.S. Court of Appeals for the Second Circuit. The Supreme Court would become involved only later—and only if it chose to hear the case.

So, no, this is not the “administrative branch” controlling copyright law. Federal agencies belong to the executive branch, and the administration is participating as an interested advocate. The court remains free to accept or reject every word of its argument.

That is separation of powers working normally, not being suspended.

The executive order does not change copyright law

The administration’s brief cites President Trump’s 2025 executive order declaring that the United States should maintain global leadership in artificial intelligence. That order establishes executive-branch policy. It directs officials to develop an AI action plan and reconsider earlier federal policies viewed as barriers to innovation.

But an executive order cannot erase rights Congress placed in the Copyright Act. The order itself says it must be implemented consistently with applicable law. It gives the administration a policy reason to favor OpenAI’s position; it does not give OpenAI a new statutory defense.

That distinction matters. “American AI leadership” may be a powerful political argument. It is not a fifth fair-use factor.

Don’t existing laws make the answer obvious?

They give us the test, but they do not predetermine the result.

Section 107 tells courts to consider four things: the purpose and character of the use, the nature of the copyrighted work, how much was used, and the effect on the work’s actual or potential market. Courts weigh those factors together and apply them to the specific conduct at issue.

That is why saying AI training is “transformative” does not end the case. OpenAI can argue that a model analyzes works to learn patterns and produces something functionally different. The Times can answer that OpenAI copied expressive journalism at enormous scale to build commercial products that can summarize, imitate, retrieve, or compete with the original reporting.

Both arguments fit inside existing copyright doctrine. The unresolved question is how that doctrine applies to the mechanics, sources, outputs, and market effects of this particular system.

Earlier cases point in competing directions. Courts have found fair use where Google digitized books to create a searchable index and where Google reused parts of a software interface for a new computing environment. But the Supreme Court’s 2023 Warhol decision warned that calling a new work “transformative” is not a magic phrase—especially when the challenged use shares a commercial purpose with the original.

The U.S. Copyright Office reached the appropriately unexciting conclusion in its AI-training report: some training uses may qualify as fair use and others may not. Lawful access, the purpose of the copying, the type of work, the model’s outputs, and harm to existing or reasonably developing licensing markets can all matter.

That is not a clean rule because Congress did not write an AI-specific one.

The Anthropic comparison is incomplete

The article points to the Anthropic litigation to suggest that courts have generally favored AI companies on training. But the crucial lesson is narrower.

A court can treat the act of training as fair use while separately treating the acquisition or retention of pirated copies as infringement. “What did the model developer do with the books?” and “How did it obtain the books?” are different legal questions.

That distinction hurts the simplistic arguments on both sides. A ruling favorable to training does not create a license to pirate. A finding that a company obtained works unlawfully does not automatically establish that every later computational use was infringing.

What actually happened here

The Trump administration has placed the political and economic power of the United States behind a broad conception of fair use for AI development. That intervention may influence the judge, and it signals how the current executive branch wants copyright doctrine to develop.

But the administration has not “defended OpenAI” in the sense of taking over OpenAI’s private defense, and it cannot decide the case. It submitted an argument permitted by federal law. OpenAI remains the defendant. The Times remains the plaintiff. The judge remains the decision-maker.

The real story is not that Trump settled the legality of AI training. The real story is that a politically consequential administration has asked the judiciary to stretch an old, deliberately flexible doctrine across a new technology—while Congress has still declined to draw a clearer line.

That may be important. It is not a ruling.

Sources

United States statement of interest

17 U.S.C. § 106—exclusive rights

17 U.S.C. § 107—fair use

28 U.S.C. § 517—United States participation in pending suits

2025 executive order on American AI leadership

U.S. Copyright Office report on generative-AI training

Google LLC v. Oracle America, Inc.