President Trump’s firing of U.S. copyright chief over Generative AI report threatens to politicize AI copyright debate

We intended to comment on the USCO’s report anyway, but before we got to do it, the news broke that Mrs. Perlmutter has been removed from office, and the timing (particularly on a weekend) makes it highly unlikely that there was any other reason than the document the agency published the day before. If anything, it could be that there already was pressure on the USCO, which justified the release of a “pre-publication version” of part 3 of its report with “congressional inquiries and expressions of interest from stakeholders.” It’s possible that some people wanted to get that document out while they still had the authority to publish it.

According to her LinkedIn profile, Mrs. Perlmutter became the Director of the USCO in October 2020, toward the end of President Trump’s first term. That means she was appointed by him, continued to serve under President Biden, and has now been fired at the beginning of President Trump’s second term.

Prior to that role, she served as Chief Policy Officer and Director for International Affairs of the United States Patent & Trademark Office (USPTO) starting in January 2012 (the month in which President Obama’s second term started). The USPTO is the more important agency, but in the end it’s a more attractive job to head a smaller agency than to have limited responsibility within a more powerful one.

The relevance of the USCO report to judicial decisions

There is a clear separation of powers in the United States: the executive government neither makes the law (which is the prerogative of Congress) nor provides binding interpretations (which is the exclusive authority of the courts). Therefore, a paper like the USCO’s AI report could even be completely ignored by the judges deciding AI copyright cases. But statements by government agencies can serve as what is called persuasive authority. They can help convince judges that a certain interpretation of the law is correct. Government agencies are independent from the parties to any particular AI copyright litigation, and no one would deny that the USCO knows a thing or two about copyright law.

In some jurisdictions such as the EU, such governmental statements that do not have legislative effect are commonly referred to as “soft law.”

There has probably never been a situation where a court wrote in its decision that a case was a wash, but what tipped the scales was a government paper on the subject. It’s more like judges will reference a government statement supporting their opinion among various other authorities, the emphasis being on statutory interpretation and judicial rulings. That, however, does not mean that government statements might not have made the difference at the stage of judicial deliberations.

There is one stage of proceeding, however, where government statements can strongly influence the course of events. Sometimes the Supreme Court of the United States, which is certain to receive various AI copyright-related petitions in the years ahead, issues a Call for the Views of the Solicitor General (CVSG), which effectively means asking the Department of Justice (DOJ) to express an opinion on whether a legal question is worthy of top court review. It would not be the USCO’s job to respond to a CVSG, but if the DOJ submitted one position and the USCO said the opposite, the DOJ’s influence could be adversely affected. There was an odd situation about six years ago with the DOJ (during President Trump’s first term) submitting an amicus curiae brief against its sister agency, the Federal Trade Commission (FTC), in the latter’s Qualcomm case.

Even a withdrawn “pre-publication” government statement can still be cited as persuasive authority. The other side will then say that the court should ignore the statement because it was ultimately never adopted. But the judges looking at this would also see that it was a decision by the Trump White House, and some judges (particularly Democratic ones) may not consider the paper less persuasive. Even the opposite could be the case. In less than two weeks, Judge William H. Alsup of the United States District Court for the Northern District of California will hear Anthropic’s motion for summary judgment (based on fair use) against a class action brought on behalf of book authors (April 4, 2025 ai fray article). That judge is clearly not a supporter of President Trump, though his decisions in Oracle v. Google did not favor the copyright holder (Oracle in that case) either.

Implications for policy-making

Whichever side loses the AI copyright court battles will try to persuade Congress to amend U.S. copyright law. And some lawmakers even want to take action before the courts have spoken. For example, Congressman Adam Schiff (D-Cal.) proposed the Generative AI Copyright Disclosure Act of 2024 (Congressional webpage), which doesn’t appear to have much traction though.

After yesterday’s decision to fire Director Perlmutter, it’s clear that the Trump Administration would advocate changing the law in favor of AI providers rather than right holders. Whether Republicans in Congress agree is not clear yet. There may also be Democrats (other than Representative Schiff) who would support the development of AI systems.

In the EU, President Trump has always been very unpopular, and he is known not be a fan of the EU. The USCO statement won’t play a role in court proceedings there, given that there are significant differences between U.S. and EU copyright law (the latter does not even have a fair use exception), but there is a risk now of anti-Trump Europeans being even more inclined to consider legislative measures favoring copyright owners over the American-dominated tech industry. That could further widen the gap between the U.S. and the EU in the technology sector in general and in AI in particular, but there definitely are people in the EU Council, the European Commission and the European Parliament who have what some of the President’s supporters call Trump Derangement Syndrome (TDS).

Why AI providers probably won’t savor the pre-publication USCO paper

It’s key to distinguish between the result and the derivation. For large parts of it, the pre-publication document comes across as balanced and nonjudgmental. Unlike an advocacy paper that supports only one side’s arguments and dismantles those of the other side (if it mentions them at all), it discusses the key points raised by both sides of the debate. But the paper does not stop there. It goes on to then express views on how the law should be applied to GenAI copyright cases, and what the commercial outcome should be.

The USCO’s pre-publication report is more favorable to the interests of copyright holders than those of AI providers in that it

• suggests that AI providers, at least those who do it on a large commercial scale and not just for research papers, need copyright licenses;
• opposes the idea that copyright owners need to opt out of the use of their material for LLM training;
• wants to give copyright holders the option to deny licenses (no compulsory licensing); and
• to the extent that copyright owners want to negotiate with AI providers through Collective Management Organizations (CMOs), the USCO suggests there is probably no antitrust issue, though it recognizes that this would be for the Antitrust Division (ATR) of the DOJ to opine on.

Fair use is the key question, and it is a multifactorial one. The USCO could have taken a right holder-friendlier position on whether the use of pirated material for LLM training precludes a finding of fair use: it merely suggests that it weighs against it. But all in all, the paper reflects a great deal of sympathy with right holders and their ecosystem. For example, the way LLMs consume copyright material is described as very distinguishable from how humans read books. And the following sentence is music in the ears of strong-copyright advocates, given that it describes the scenario at issue in all of the major GenAI copyright cases currently pending in the U.S.:

“[M]aking commercial use of vast troves of copyrighted works to produce expressive content that competes with them in existing markets, especially where this is accomplished through illegal access, goes beyond established fair use boundaries.”

Also, the USCO paper potentially supports the notion that AI providers infringe copyright at different stages of the value chain, from training to output.

The penultimate paragraph sounds balanced, but implicitly denies that AI model training constitutes fair use:

“In our view, American leadership in the AI space would best be furthered by supporting both of these world-class industries that contribute so much to our economic and cultural advancement. Effective licensing options can ensure that innovation continues to advance without undermining intellectual property rights. These groundbreaking technologies should benefit both the innovators who design them and the creators whose content fuels them, as well as the general public.”

What other options than replacing the head of the USPTO could have helped diminish the influence of the paper?

Firing Director Perlmutter may turn out counterproductive. In the end, judges and lawmakers would have understood that people who dedicate their career to IP law are naturally predisposed to favor what gives their field more influence. That doesn’t mean they all advocate broader rights and stronger enforceability, but compared to the average person, those experts are more receptive to whatever makes their field of law more important. There are exceptions to the rule, but it works as a rule of thumb.

The Trump Administration could have had a different agency issue a statement on AI and copyright that would have been tech-friendlier, either proactively or whenever there would have been an opportunity to file an amicus curiae brief in one of the ongoing lawsuits. In that case, there would have been an intragovernmental disagreement on particular questions, but with what happened on Friday (release of pre-publication paper) and Saturday (firing of Director Perlmutter), it’s equally clear that there are different views in the federal government. The next USCO Director can align that office with the policy goals of the White House, but what’s been published has been published, even if it was labeled “pre-publication.”

Any of that would also have avoided a debate over the legality of the dismissal. A few days earlier, President Trump had fired the Librarian of Congress, who is right above the head of the USCO. The top Democrat on the Commitee on House Administration, Rep. Joe Morelle from New York, issued a statement yesterday that calls it “a brazen, unprecedented power grab with no legal basis” and blames it all on “Elon Musk’s efforts to mine troves of copyrighted works to train AI models,” though Mr. Musk does not get support for all of his business interests. He is not in favor of tariffs, and is not involved with Project Stargate (January 23, 2025 ai fray article).

The focus will soon shift again to the various pending cases. They’re all still at a stage where any fair use decisions will be appealable.