ARTICLE
12 February 2025

Copyright Office Publishes Part 2 Of Its Artificial Intelligence And Copyright Report

On January 29, 2025, the U.S. Copyright Office (USCO) published Part 2 of its Artificial Intelligence (AI) and Copyright Report. While Part I, released on July 31, 2024...
United States Intellectual Property

On January 29, 2025, the U.S. Copyright Office (USCO) published Part 2 of its Artificial Intelligence (AI) and Copyright Report. While Part I, released on July 31, 2024, focused on “deepfakes” (realistic digital representations of individuals' voices or appearances), Part 2 addresses the copyright status of creative works produced through AI. Specifically, it provides some clarity and guidance on how much human input is required for AI-assisted works to be deemed eligible for copyright.

1. Human Authorship Remains Essential. First and foremost, the report reiterates the longstanding copyright principle that only human-made works can receive copyright protection. Consequently, any work solely created by an AI system is ineligible for copyright unless there is a demonstrable level of human direction and influence over the final product. Courts have consistently upheld that copyright protection in the U.S. requires human authorship, based on the Copyright Clause of the Constitution and judicial interpretation of copyright law. It explains that while using AI as a tool does not negate copyright protection, the resulting work is only copyrightable if it contains sufficient human-authored expressive elements.

Who would have thought we'd be asking the question “how much human do we need?” in any legal analysis but here we are. There are many different kinds of human contributions in AI generated outputs including prompts, expressive inputs, and modification and arrangement of AI outputs. Interestingly the report states that prompts alone do not currently provide sufficient human control to make users the authors of artificial intelligence outputs. However, human-authored expressive inputs that are perceptible in the output may be protected. Additionally, creative selection, arrangement, or modification of AI-generated content can be copyrightable.

2. Using AI as a Tool or as the Creator. Part 2 makes it very clear that there is a difference between using AI as a tool or as the creator itself. For example, the USCO explains that AI may be used for various supportive tasks, such as idea generation, image refinement, or grammatical checks but it is the substantial creative contribution from a human author that qualifies a work for copyright. If AI autonomously generates a piece with no meaningful human direction or modification, it remains outside the scope of copyright.

3. Good Output Does not Equal Good Outcome. According to the USCO, even the most detailed or elaborate text prompt does not secure copyright ownership over the resulting AI-generated material. Essentially, prompts function as instructions rather than creative expressions in their own right. While a prompt might convey an idea, the AI system applies its own algorithms and training data—beyond the user's direct control—to create the final output. The AI system essentially takes control of execution, using its algorithms and training data in ways not fully dictated by the user, which makes it impossible to designate the user as the work's sole author.

However, the USCO also notes that “where a human inputs their own copyrightable work and that work is perceptible in the output, they will be the author of at least that portion of the output.”

4. International Standards. The report surveys international approaches to AI as well, noting that most agree copyright requires human authorship. However, interpretations and applications of copyrightability standards for AI-generated works are still evolving internationally. Some commentators advocate for updated legal frameworks to accommodate emerging AI technology, but the USCO concludes that “existing legal doctrines are adequate and appropriate” for deciding whether AI outputs warrant copyright, citing copyright law's historical adaptability to new technologies. The agency also underscores that it will keep tracking new artificial intelligence tools and legal developments to ascertain if a different strategy becomes necessary.

5. It's a Case-by-Case Analysis. This phrase has saved many a lawyer but it's because it's an accurate answer. There is no hardline so each case does have to be examined in its totality. AI works can still qualify for copyright protection if a human engages sufficiently but what that means will depend on many factors. Human engagement can include editing, arranging, or selecting AI-created elements or even utilizing AI outputs into a larger piece that was predominantly authored by humans. The Office emphasizes that each scenario must be examined individually to determine whether the human involvement was substantial enough to meet the copyright threshold, hence the “case by case” analysis.

While there are many more questions surrounding the use of AI, this report does offer up some very practical advice when it comes to the specific issue of copyrightability as it relates to works incorporating the use of artificial intelligence. Companies and users of AI should:

  • Ensure that there is in fact meaningful human creativity in projects that use AI tools. Track where and when AI is used and document to what extent humans contributed, since evidence of human authorship will likely be essential for future copyright claims.
  • Modify AI outputs with genuine original material.
  • Do your due diligence when it comes to using AI tools and address questions of ownership and legal liability before distributing, selling, or licensing such works.
  • Regularly update guidance and practices.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.

Mondaq uses cookies on this website. By using our website you agree to our use of cookies as set out in our Privacy Policy.

Learn More