Analyzing critical legal trends and developments across data, cyber, AI and digital regulations from around the world and beyond borders

On January 29, 2025, the United States Copyright Office published its second installment in a series of reports on Copyright and Artificial Intelligence (the “Report”), focusing on issues of copyrightability posed by works generated by or with the assistance of AI tools.

Background

The Report is the culmination of a years’ long exploration of the topic by the Copyright Office. In 2022, the Office refused to register a work of visual art that the applicant claimed was created autonomously by an AI program—a decision subsequently upheld by District Court for the District of Columbia. The following year, the Office allowed registration of a comic book that included images generated by AI, explaining that the protection only applied to the human-authored components of the work. To elucidate it official position, the Copyright Office issued March 2023 guidance on the registration of works containing AI-generated materials, instructing applicants to specify the human contributions of works containing more than a de minimis amount of AI-generated content.

Later in 2023, the Copyright Office issued a Notice of Inquiry seeking public comment on a variety of questions regarding the implications of AI technologies on copyright law, receiving over 10,000 responses. The Copyright Office published a first report addressing copyright issues pertaining to deepfakes in July 2024.

The Report largely reiterates and builds on Copyright Office policies and the recent decisions, emphasizing the indispensability of human creativity to obtaining copyright protection. The report, however, differs markedly from the earlier report on deepfakes, which noted various shortcomings in the existing legal frameworks and called for new legislation; the Report on copyrightability states that existing laws are sufficiently flexible to address the emerging issues posed by AI and no legislative changes are required.

In Detail

Use of Generative AI to Assist Creativity: The Report confirms that using AI as a mere aid in the creative process does not preclude copyright attaching to resulting works. Examples cited in the Report include the use of AI to generate ideas, structure a preliminary outline for literary works, “age” or “de-age” actors, identify chord progressions, detect errors in software code, and remove unwanted objects from a scene. The Report draws a distinction between using AI as a stand-in for human creativity—which would be suspect under copyright laws requiring a threshold level of original expression—and leveraging AI to assist in the creation of works.

User Prompts: Questions regarding the copyright treatment of user prompts to an AI system attracted particular attention from the public. Commenters generally agreed that a prompt itself can be subject to copyright protection assuming it meets the other requirements for copyright, like originality. There was less consensus as to whether the author of a prompt could claim copyright in the resulting AI output. The Report concludes that prompts alone do not give the user of an AI system sufficient control over the resulting output (at least with respect to the current state of the art) to confer copyright on the author of the prompt. Prompts are instead more akin to instructions that convey unprotectable abstract ideas—similar to a would-be joint author who merely suggests ideas to a collaborator, who then does the work of converting those concepts into copyrightable expression. The Report expressly contemplates that as the technology progresses to allow greater user control over the output may affect this conclusion, though advancements geared towards greater automation would strengthen the Report’s arguments that a prompt’s author has no protectible copyright interest in the resulting AI output.

Expressive Inputs: The Report differentiates between prompts that simply indicate a desired output and “expressive inputs” into a generative AI system, giving the example of a user that uploads an illustration into an AI system and asks the system to make certain alterations to the image. The Report explains that where the input itself constitutes a copyrightable work and there are protectable elements of that work that can be recognized in the output, the author can claim copyright in the output at least as concerns those recognizable features.

Modifying AI Content: The Report further explains that an author may use an AI-created output as an initial or intermediate step, modifying or arranging the output into a final creative work that itself is the product of human creativity. The Report provides the example of the combination of human-authored text with AI-generated  images in a comic book, which is protectible as a compilation work. The Report also notes that some generative AI tools even offer features that allow users to manipulate, alter, or arrange works generated by the tool into a distinct final product.

Global Approaches: The Report acknowledges that many other jurisdictions are also grappling with issues of copyrightability in AI-generated content. While most have arrived at the same conclusion that copyright requires human authorship, they have taken different approaches in applying this precept. For example, in Japan the Cultural Council has developed a four factor test for determining whether an AI-generated work is copyright eligible. A Beijing court found that an image created by an AI model was protected, citing the fact that the user selected from over 150 prompts and made modifications to the AI-created image. The UK has considered whether to amend its copyright statute, which provides protection for works “generated by computer in circumstances such that there is no human author of the work,” designating the “person by whom the arrangements necessary for the creation of the work are undertaken” as the author.

Legislative Changes: The Report concludes by considering possible changes to the Copyright Act pertaining to the availability of protection for AI-generated or -assisted works. While some commenters suggested that conferring protection on AI works—either by expanding the scope of copyright or by created a separate sui generis right—the majority shared the view of the Copyright Office that copyright should be limited to human creations. The Report also observed that the extension of copyright protection is unnecessary, because numerous incentives already exist to promote innovation in the development of generative AI models. The Report also acknowledged the view that generative AI tools may enable people with physical or cognitive impairments to engage in creative activities that would otherwise be foreclosed to them and copyright should encourage such work.  The Report considered that these instances were adequately addressed insofar as they may constitute works where AI is used to assist, rather than substitute for, human creation and therefore would be copyright-eligible.

Next Steps

As noted above, the Report hews closely to established Copyright Office policies and laws applicable to the copyrightability of AI-generated materials and declines to recommend changes to existing laws. While the Report does not represent a sea change in copyright policy, the Report nevertheless contains many insights into how the Copyright Office views works generated with the assistance of AI. Organizations that use AI to generate materials or to assist with coding, should take particular note and should document, internally, the ways that AI is used and consider Copyright Office guidance when creating or revising internal policies on IP identification and rights management.

Given the limitations regarding the copyright protectability of certain aspects of AI as outlined in the Report, it continues to be important that providers and users of AI take a holistic approach to protecting their AI work, including, for example, internal governance and protections, evaluating the availability of other IP rights such as trade secrets, and paying careful attention to contracting for AI and the allocation of rights in such contracts.

Although the Report largely maintains the status quo, it provides an illustrative breakdown of common AI use cases, including those that may result in copyright protection.  The intersection of copyright and AI is set to be the subject of continual discussion and as a reminder, the third (and final) report in the series, expected in the first quarter of 2025, will cover the training of AI models on copyrighted works, licensing considerations, and allocation of liability.

Author

Cynthia J. Cole is Chair of Baker McKenzie’s Global Commercial, Tech and Transactions Business Unit, a member of the Firm’s global Commercial, Data, IP and Trade (CDIT) practice group steering Committee and Co-chair of Baker Women California. A former CEO and General Counsel, just before joining the Firm, Cynthia was Deputy Department Chair of the Corporate Section in the California offices of Baker Botts where she built the technology transactions and data privacy practice. An intellectual property transactions attorney, Cynthia also has expertise in AI, digital transformation, data privacy, and cybersecurity strategy.

Author

Adam Aft helps global companies navigate the complex issues regarding intellectual property, data, and technology in product counseling, technology, and M&A transactions. He leads the Firm's North America Technology Transactions group and co-leads the group globally. Adam regularly advises a range of clients on transformational activities, including the intellectual property, data and data privacy, and technology aspects of mergers and acquisitions, new product and service initiatives, and new trends driving business such as platform development, data monetization, and artificial intelligence.

Author

Avi Toltzis is a Knowledge Lawyer in Baker McKenzie's Chicago office.