Charting a Course on AI Policy: the U.S. Copyright Office Speaks! 

By Gregory J. Krabacher, Epstein Becker Green

Recently, the U.S. Copyright Office published the second of an intended three-part report entitled “Copyright and Artificial Intelligence.” Here are those three parts:

• Part 1: Digital Replicas (aka “Deep Fakes”), published July 31, 2024;¹

• Part 2: Copyrightability, published January 29, 2025;² and

• Part 3: Legal implications of training AI models on copyrighted works, licensing considerations, and the allocation of any potential liability, publication date to be decided.³

Collectively, this report – I’ll refer to it here as the “AI Report” –  is intended to provide the Copyright Office’s perspective on fundamental questions for copyright law and policy implicated by a technology described by the Office as presenting both “exciting opportunities” and potentially “existential” threats.

As framed by Shira Perlmutter, Register of Copyrights and Director U.S. Copyright Office, in her forward to publication of Part 1 of the AI Report, these fundamental questions include:

To what extent will AI-generated content replace human authorship? How does human creativity differ in nature from what AI systems can generate, now or in the future? What does this mean for the incentive-based foundation of the U.S. copyright system? In what ways can the technology serve as a valuable tool to amplify human creativity and ultimately promote science and the arts? How do we respect and reward human creators without impeding technological progress?⁴

The AI Report builds on the Copyright Office’s intensive focus on AI issues in recent years.  As previously reported in ILN IP Insider,⁵ the Copyright Office published guidance on March 12, 2023, entitled “Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence.”  This document continues to guide Copyright Office staff and is a must read for any copyright lawyer practicing before the Office.

Later that same year, on August 30, 2023, the Office issued a notice of inquiry (NOI) requesting comments on copyright law and policy issues arising from the development and use of AI systems.⁶  More than 10,000 responses were received from a broad range of perspectives, including comments from all 50 U.S. States and 67 countries.  Notably, the Copyright Office’s work in this area pre-dates President Biden’s sweeping Executive Order No. 14110 on the topic of Artificial Intelligence and continues after the revocation of this Order by the next administration.⁷

U.S. Copyright Office AI Report Part 1: Digital Replicas (aka “Deep Fakes”)

As noted in the New Years Resolutions article appearing here in January, if there is one area of low-hanging fruit in which bi-partisan agreement can be reached for new legislation on AI, deep fakes may well be it.⁸  Part 1 of the AI Report adds further impetus for pursuing such legislation, concluding that a new federal law is urgently needed.  Interestingly, however, the Copyright Office recommends that any new federal law supplement, but not preempt, the existing state laws regarding deep fakes.  As with similar proposals for U.S. comprehensive privacy legislation, preemption is sure to be a hotly debated issue with substantial interest from industry stakeholders.

In Part 1 of the AI Report, the Office extensively reviews existing U.S. federal and state laws, finding that the current patchwork of state laws, federal agency enforcement, and lack of a unifying federal standard, fails to adequately or consistently address the harms caused by deep fakes.  The report also includes a review of recent Congressional activity, including federal legislation introduced to address unauthorized digital replicas in various contexts, such as political⁹ advertisements and communications and sexually explicit images.  However, Part 1 misses the opportunity to incorporate analysis of international approaches to this topic.  For example, Article 50 of the EU AI Act (finalized just prior to the Copyright Office’s Report but pending in various forms for several years prior¹⁰), specifies a framework in which deployers of deep fake content are required to clearly disclose that it has been artificially created or manipulated.¹¹

Part 1 of the AI Report concludes with an interesting analysis of the issues arising from AI “outputs that imitate the artistic style of a human creator.”¹²  The Office notes that while imitation (or “inspiration”) has long been part of the creative process in many artistic disciplines, the new issues created by AI derive mainly from the speed and scale at which this imitation can be achieved.  This results in real concerns for the market for the original artist’s work.  Indeed, the Report notes that some artists are reportedly finding themselves competing against their distorted AI replicas.  While recognizing the seriousness of these issues, the Copyright Office concludes that U.S. copyright law is limited in its application to these areas, some of which may be better handled under other sources of law, such as privacy, rights of publicity, passing off, and unfair and deceptive trade practices.  This conclusion is curious in the context of the rest of the report, which argues at length (and with merit) that these same disparate sources of law result in inconsistent outcomes with respect to the other kinds of harm arising from the use of deep fakes.

It remains to be seen whether the specific proposals of the Copyright Office are adopted in a new federal law, or, for that matter, whether any new federal law at all will be enacted.  In the meantime, the specific proposals put forth by the Office will provide guidance for federal and state legislatures.

Without further ado, here now are the main points of the Copyright Office’s proposal for a new federal statute on digital replicas:¹³

• Subject Matter. The statute should target those digital replicas, whether generated by AI or otherwise, that are so realistic that they are difficult to distinguish from authentic depictions. Protection should be narrower than, and distinct from, the broader “name, image, and likeness” protections offered by many states.

• Persons Protected. The statute should cover all individuals, not just celebrities, public figures, or those whose identities have commercial value. Everyone is vulnerable to the harms that unauthorized digital replicas can cause, regardless of their level of fame or prior commercial exposure.

• Term of Protection. Protection should endure at least for the individual’s lifetime. Any postmortem protection should be limited in duration, potentially with the option to extend the term if the individual’s persona continues to be exploited.

• Infringing Acts. Liability should arise from the distribution or making available of an unauthorized digital replica, but not the act of creation alone. It should not be limited to commercial uses, as the harms caused are often personal in nature. It should require actual knowledge both that the representation was a digital replica of a particular individual and that it was unauthorized.

• Secondary Liability. Traditional tort principles of secondary liability should apply. The statute should include a safe harbor mechanism that incentivizes online service providers to remove unauthorized digital replicas after receiving effective notice or otherwise obtaining knowledge that they are unauthorized.

• Licensing and Assignment. Individuals should be able to license and monetize their digital replica rights, subject to guardrails, but not to assign them outright. Licenses of the rights of minors should require additional safeguards.

• First Amendment Concerns. Free speech concerns should expressly be addressed in the statute. The use of a balancing framework, rather than categorical exemptions, would avoid overbreadth and allow greater flexibility.

• Remedies. Effective remedies should be provided, both injunctive relief and monetary damages. The inclusion of statutory damages and/or prevailing party attorney’s fees provisions would ensure that protection is available to individuals regardless of their financial resources. In some circumstances, criminal liability would be appropriate.

• Relationship to State Laws. Given well-established state rights of publicity and privacy, the Office does not recommend full federal preemption. Federal law should provide a floor of consistent protection nationwide, with states continuing to be able to provide additional protections. It should be clarified that section 114(b) of the Copyright Act does not preempt or conflict with laws restricting unauthorized voice digital replica.

U.S. Copyright Office AI Report Part 2: Copyrightability

In Part 2 of the AI Report, the Copyright Office provides its view on the copyrightability of outputs of AI systems.  In a divergence from the Office’s advocacy for new legislation in the specific area of deep fakes addressed in Part 1, the Office recommends a slower course on the topic of copyrightability in Part 2.  This is unsurprising, given the need for a careful, balanced, and narrowly focused approach to the regulation of this nascent technology.  As we wrote in January:

Although a range of IP-related issues have been raised by recent developments in AI technology and its applications, some are not yet ripe for government intervention. It will be vital to avoid overreach and understand the potential costs and benefits as much as possible. Any new IP-related legislation or regulations should target specific known issues or problems; tailor definitions, requirements, and consequences narrowly; reduce uncertainty rather than increase it; and focus on improving the ability of the private sector to innovate and creators to thrive.¹⁴

Furthermore, the Copyright Office notes that all other countries to have considered the matter have concluded, consistent with the Copyright Office’s own view, that copyright requires human ownership.  According to the Office, this includes Korea, Japan, China, the European Union, the United Kingdom, Hong Kong, India, New Zealand, Canada, and Australia.¹⁵  Striking a cautionary note, the Office finds that although some level of consensus on the need for human authorship appears to be emerging, and most countries have so far continued to apply existing law, it is clear that views are still being formed. Given that it remains to be seen how copyrightability standards will be interpreted and applied, the Office indicates its intent to closely monitor developments abroad and evaluate how other countries’ evolving approaches may ultimately overlap or differ from those in the U.S.  As we have likewise noted, harmonization of approach on an international level is an important goal.¹⁶

In particular, the Office basis its recommendation for a “wait and see” approach on the following points, as summarized in the report:¹⁷

• Questions of copyrightability and AI can be resolved pursuant to existing law, without the need for legislative change.

• The use of AI tools to assist rather than stand in for human creativity does not affect the availability of copyright protection for the output.

• Copyright protects the original expression in a work created by a human author, even if the work also includes AI-generated material.

• Copyright does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements.

• Whether human contributions to AI-generated outputs are sufficient to constitute authorship must be analyzed on a case-by-case basis.

• Based on the functioning of current generally available technology, prompts do not alone provide sufficient control.

• Human authors are entitled to copyright in their works of authorship that are perceptible in AI-generated outputs, as well as the creative selection, coordination, or arrangement of material in the outputs, or creative modifications of the outputs.

• The case has not been made for additional copyright or sui generis protection for AI-generated content.

Concluding Thoughts

The current appetite for, and direction of, new AI legislation at the federal level in the U.S. is presently unclear.  Reportedly, the change in administration has resulted in a shift in focus and strategy, both within government and in the private sector – focusing more on growth and development of the AI sector and less on regulations and safety concerns.¹⁸  While, as a general matter, such a change in focus could well result in a different direction for the development of an evolving U.S. legal framework related to AI, the particular issues addressed by the Copyright Office in the AI Report have continuing resonance.  The balanced and thoughtful analysis found in the U.S. Copyright Office’s AI Report is a resource for Congress, the courts, and the executive branch in furthering the Constitutional goal “to promote the progress of science and useful arts…”¹⁹  We look forward to Part 3 of the AI Report in the coming months and to the evolving debate on these important issues.