On Saturday, the Trump administration fired Shira Perlmutter, the register of copyrights and director of the U.S. Copyright Office, just two days after the dismissal of Librarian of Congress Carla Hayden, under whose auspices the U.S. Copyright Office operates. Perlmutter was appointed by Hayden in 2020.

By law, the register works under the direction of the Librarian of Congress and carries out a variety of legal and policy functions specified in Title 17 of the United States Code. The Copyright Office employs approximately 450 staff members, including four associate registers of copyrights and two assistant registers of copyrights, who report directly to the register.

The move, like Hayden’s dismissal before it, was soon blasted by Democratic members of Congress and industry figures. Rep. Joe Morelle (Dem., N.Y.), the top Democrat on the Committee on House Administration, called the move “a brazen, unprecedented power grab with no legal basis,” adding, “It is surely no coincidence he acted less than a day after she refused to rubber-stamp Elon Musk’s efforts to mine troves of copyrighted works to train AI models.”

A letter sent to Rep. Bryan Steil (R, Wis.), chair of the Congressional Joint Committee on the Library, on May 14, organized by the eBook Study Group and signed by organizations including EveryLibrary Institute and Library Futures, also took the move to task.

“The U.S. Constitution charges Congress with the responsibility ‘to promote the Progress of Science and useful Arts’ by granting limited rights to creators, rights designed to advance knowledge, creativity, and the public good. President Trump’s abrupt dismissal of the Register, preceded by his termination of the Librarian of Congress, represents a serious and unprecedented intrusion by the Executive Branch into the Legislative Branch’s domain and authority,” the letter reads. “This is not just a staffing change; it is a constitutional disruption. Such interference threatens the independence and credibility of the Copyright Office and sets a dangerous precedent that endangers the integrity of the entire copyright system.”

On May 9, Perlmutter’s office released the pre-publication version of the third part of the Copyright Office’s report “Copyright and Artificial Intelligence,” following a first segment released in July 2024 and a second released this January. This third part focuses on the impact of generative AI training across a wide range of topics, from the origins of the technology, to AI companies’ possible infringement in training their data sets, to those companies’ defense that the training counts as fair use, to what potential licensing scenarios might look like.

Morelle and others have speculated that Perlmutter’s dismissal was likely due to her release of the preliminary report. But sources close to the office, who spoke with PW on condition of anonymity, suggest that it is more likely that Perlmutter, having heard of her impending dismissal, ordered the report released beforehand to ensure it entered the public record in spite of its incomplete status. (The report, for instance, lacks some citations.)

While the report is not a legally binding document, its recommendations on and to AI companies have not been taken well by tech industry titans. Perlmutter’s dismissal comes as the Trump administration continues to cozy up to the tech industry, whose leaders have consistently criticized the U.S. Copyright Office’s role in issuing guidance on copyright and intellectual property law, especially as it relates to artificial intelligence. On April 11, Jack Dorsey, founder of Twitter (now X), tweeted, simply, “Delete all IP law.” Current X owner Elon Musk, who oversees the Trump administration’s extra-governmental consultative body known Department of Government Efficiency, replied, “I agree.”

In March, the powerful Silicon Valley investment firm Andreessen Horowitz, which has some $45 billion under management, wrote in response to a request for information on the development of an AI action plan. “First, neither the Copyright Office nor any other government agency should release guidance related to this issue—or other issues critical to American competitiveness in AI—until the conclusion of the National AI Action Plan process,” read the statement, referring to a February 25 executive order issued by president Trump. “Any rulemaking, report, guidance, or policy statement that was prepared by prior administration appointees is unlikely to reflect the current administration’s policy preferences.”

Meanwhile, many AI companies are lobbying the government directly for more access to data held or supported by the government in order to train their generative models. OpenAI justified such requests in a March memo to the U.S. Office of Science and Technology Policy by insisting that additional access “would boost AI development in any case, but would be particularly important if shifting copyright rules restrict American companies’ access to training data.” The implication is that technology companies hope to push the government to share access to its data as quickly as possible in order to preemptively circumvent any future legal decisions that might compel the companies to license the data or otherwise deny them access.

The Copyright Office’s report explicitly states that “the copying involved in AI training threatens significant potential harm to the market for or value of copyrighted works,” adding: “Where a model can produce substantially similar outputs that directly substitute for works in the training data, it can lead to lost sales. Even where a model’s outputs are not substantially similar to any specific copyrighted work, they can dilute the market for works similar to those found in its training data, including by generating material stylistically similar to those works.”

Still, the report strives to be even-handed: “The assessment of market harm will also depend on the extent to which copyrighted works can be licensed for AI training. Voluntary licensing is already happening in some sectors, and it appears reasonable or likely to be developed in others—at least for certain types of works, training, and models.”

Maria Pallante, president and CEO of the Association of American Publishers, shared her own thoughts on the report with PW. “Speaking as a former register, the measure of whether the Copyright Office is performing a neutral, non-partisan, expert analysis is that no stakeholder will be entirely happy with its opinion. That is the case here, too,” she wrote in an email. “There is plenty in its analysis for the creative sector to appreciate and some perspectives that are more helpful to AI developers—I think content owners will strongly disagree with some of the transformative use analysis related to coding, for example, but the market dilution arguments for copyright owners is innovative and spot-on. But the report is a resource—lawmakers and courts will embrace or distinguish it.”

This article has been updated with further information.