# US Copyright Office 2025 Report

The US Copyright Office 2025 Report is the final report on copyright and artificial intelligence, released in 2025, addressing legal and policy questions around AI-generated works.

The United States Copyright Office (USCO) released its final report on copyright and artificial intelligence in 2025. The report is the culmination of a multi-year inquiry into how copyright law should apply to works created with or by [generative-ai](https://www.wikiprompt.org/wiki/generative-ai) systems. It addresses questions about authorship, infringement liability, and the use of copyrighted material in training [artificial-intelligence](https://www.wikiprompt.org/wiki/artificial-intelligence) models. The report provides guidance to Congress, courts, and the public on the Office's interpretation of existing law, while also recommending potential legislative changes.

The Copyright Office, a part of the Library of Congress, is the U.S. government body that registers copyright claims, records ownership information, and advises Congress on copyright policy. The 2025 report follows a series of earlier notices and public consultations initiated in 2023, which sought input from creators, technology companies, and legal experts. The final report synthesizes those comments and outlines the Office's position on several contested issues.

## Background and Process

The Copyright Office began its AI inquiry in early 2023, prompted by the rapid proliferation of generative AI tools that can produce text, images, music, and other creative works. The Office issued a notice of inquiry in August 2023, requesting public comments on a range of questions, including whether AI-generated outputs can be protected by copyright, who should be considered the author, and how existing doctrines like fair use apply to AI training.

The Office received over 10,000 comments from individuals and organizations, including major technology firms like [openai](https://www.wikiprompt.org/wiki/openai), [anthropic](https://www.wikiprompt.org/wiki/anthropic), and [google-deepmind](https://www.wikiprompt.org/wiki/google-deepmind), as well as creative industry groups, academics, and individual artists. In 2024, the Office held public listening sessions and released several discussion drafts on specific topics, such as digital replicas and the use of copyrighted works in training data. The final report, released in 2025, consolidates these findings and sets forth the Office's official recommendations.

## Key Findings on Authorship

A central question addressed in the report is whether AI-generated works can be copyrighted. The Copyright Office maintains that copyright protection requires human authorship. Under current law, works created entirely by an AI system, with no human creative input, are not eligible for registration. However, works that involve sufficient human creative control over the AI's output may qualify, depending on the degree of human involvement.

The report clarifies that the use of an AI tool as an aid does not necessarily bar copyright if a human makes creative decisions, such as selecting prompts, arranging outputs, or modifying the final result. The Office cites its earlier decisions, including the 2023 ruling that denied registration for a comic book with AI-generated images because the human contribution was deemed insufficient. The report emphasizes a case-by-case analysis, focusing on whether the human author exercised "creative control" over the expressive elements.

The report also discusses the concept of "authorship" in the context of [machine-learning](https://www.wikiprompt.org/wiki/machine-learning) models. It rejects the idea that an AI system itself can be an author, aligning with longstanding legal precedent that copyright vests in human creators. The Office recommends that Congress consider clarifying legislation to address borderline cases, but it stops short of proposing a blanket rule for all AI-assisted works.

## Training Data and Fair Use

Another major section of the report addresses the use of copyrighted works to train AI models. Many AI systems, including [large-language-model](https://www.wikiprompt.org/wiki/large-language-model)s, are trained on vast datasets that include books, articles, images, and other copyrighted material. The report examines whether such use constitutes copyright infringement or falls under the fair use doctrine.

The Copyright Office concludes that the question of fair use for AI training cannot be answered categorically. It depends on factors such as the nature of the use, the amount of material used, and the potential market impact. The report notes that some training uses may be transformative, particularly when the AI is used for research or non-commercial purposes, but that commercial uses that compete with the original works are less likely to be fair.

The report recommends that Congress consider a statutory licensing scheme for AI training, which would allow companies to use copyrighted works in exchange for compensation to rightsholders. It also suggests that courts continue to apply the four-factor fair use test on a case-by-case basis, as they have done in early litigation involving AI companies.

## Liability for AI-Generated Infringing Content

The report also addresses the liability of AI developers and users when an AI system produces content that infringes on existing copyrights. It distinguishes between the AI model itself, which is a tool, and the human user who directs its output. The Office argues that users who intentionally prompt an AI to reproduce copyrighted material may be liable for infringement, similar to a human who copies a work.

For developers, the report considers whether they can be held liable for infringing outputs generated by their models. It discusses doctrines such as contributory infringement and vicarious liability, noting that liability may attach if a developer knows that the AI is likely to produce infringing content and fails to take reasonable steps to prevent it. The report encourages developers to implement technical measures, such as filters and [model-pruning](https://www.wikiprompt.org/wiki/model-pruning) techniques, to reduce the risk of generating infringing material.

The report also examines the role of [data-augmentation](https://www.wikiprompt.org/wiki/data-augmentation) and other training practices that might inadvertently lead to memorization of copyrighted works. It suggests that developers should document their training data and maintain records to facilitate legal compliance.

## Digital Replicas and Voice Likeness

A significant portion of the report deals with digital replicas - AI-generated imitations of a person's voice, image, or likeness. The Copyright Office recognizes that existing copyright law provides limited protection against unauthorized digital replicas, as copyright does not cover a person's identity or likeness. The report recommends new federal legislation to address this gap, creating a right of publicity that would protect individuals from commercial exploitation of their digital replicas.

The report notes that some states already have publicity rights laws, but these vary widely and do not provide uniform protection. It proposes a federal statute that would establish a clear framework for consent, licensing, and remedies. The Office also suggests that the Copyright Office should maintain a registry of authorized digital replicas to help platforms and users identify legitimate uses.

## International Perspectives

The report places the U.S. approach in an international context, noting that other countries are also grappling with similar issues. It highlights the European Union's AI Act, which includes transparency requirements for AI training data, and the United Kingdom's copyright exceptions for text and data mining. The report argues that the U.S. should maintain its leadership in AI innovation while ensuring that copyright law adapts to new technologies.

The Copyright Office emphasizes the importance of international cooperation, as AI systems often operate across borders. It recommends that the U.S. engage with international bodies like the World Intellectual Property Organization to develop consistent standards for AI and copyright.

## Recommendations and Next Steps

The 2025 report concludes with a set of recommendations for Congress, courts, and the Copyright Office itself. These include:

- Clarifying that human authorship is required for copyright protection, but providing guidance on when AI-assisted works qualify.
- Enacting a federal digital replica statute to protect individuals' voice and likeness.
- Considering a licensing framework for AI training data, with input from stakeholders.
- Encouraging courts to apply existing fair use doctrine flexibly, while recognizing the unique aspects of AI.
- Updating the Copyright Office's registration practices to handle AI-related applications, including a requirement to disclose the use of AI in the creation process.

The report also announces that the Copyright Office will issue updated guidance for registration applicants, building on its 2023 statement that required disclosure of AI-generated content. The Office plans to develop new training materials for its examiners and to continue monitoring technological developments.

## Reception and Impact

The report has been met with mixed reactions. Technology companies, including [openai](https://www.wikiprompt.org/wiki/openai) and [anthropic](https://www.wikiprompt.org/wiki/anthropic), welcomed the report's nuanced approach to fair use, but expressed concerns about potential licensing requirements. Creative industry groups praised the report's emphasis on human authorship and its call for digital replica legislation, but some argued that the fair use analysis was too permissive.

Legal scholars have noted that the report does not have the force of law, but it is likely to influence courts and lawmakers. The Copyright Office's interpretations are often given deference by courts, and its recommendations may shape future legislation. The report is expected to be a key reference in ongoing litigation involving AI companies, such as the lawsuits filed by authors and artists against major AI developers.

The report also has implications for the Copyright Office's own operations. In 2025, the Office faced an unusual leadership dispute, with President Donald Trump claiming to have fired Register Shira Perlmutter and appointed Paul Perkins as acting register. However, Perlmutter's position as an employee of Congress was endorsed by courts, and the report was issued under her leadership. The dispute did not affect the substance of the report, but it highlighted the Office's unique position within the legislative branch.

Overall, the 2025 report represents a significant milestone in the ongoing effort to align copyright law with the realities of artificial intelligence. It provides a comprehensive framework for addressing the most pressing questions, while leaving room for further evolution as technology advances.

---
Source: https://www.wikiprompt.org/wiki/us-copyright-office-2025-report
License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/)
Last updated: 2026-09-13T03:52:00.326767+00:00