Who Owns AI Generated Video and What the Law Currently Says

What the published guidance actually establishes about copyright in AI assisted work, what remains unsettled, and how a commercial project should be structured in response.

Ownership of AI assisted work is the question clients ask third, after cost and speed, and it deserves a careful answer rather than a reassuring one. The position is clearer than it was and it is not fully settled, and the practical response is to structure projects so that the answer matters less.

The clearest published guidance comes from the U.S. Copyright Office, which examined the question across a series of reports. On copyrightability it concluded that existing law protects human authored expression in works created with AI tools, that outputs lacking meaningful human creative input do not qualify for protection, and that prompt selection alone, however detailed or iterative, does not by itself produce a copyrightable work.

The practical reading for a commercial project is that authorship is established by what the humans did rather than by what the tool produced. A film built through directed art direction, storyboarding, shot selection, compositing, editing, grading and sound design carries substantial human authorship. A clip pulled from a single prompt with no further intervention carries very little, which means it may be difficult to protect against reuse.

This has a direct consequence for how work should be commissioned and recorded. A studio that documents its creative process, the treatment, the visual language specification, the storyboard, the selection decisions, the compositing and the grade, is documenting the human contribution that the guidance identifies as the basis of protection. Keeping that record in the project archive is inexpensive and is what supports the position if it is ever tested.

The Office also addressed digital replicas in a separate report, examining the questions raised by AI generated likenesses of real individuals. For a studio the operational implication is unambiguous: reproducing a specific person's appearance or voice requires their informed, documented and revocable consent, specifying the uses and the duration, and this should be handled with the same seriousness as a talent buyout.

A further report addressed generative AI training, which is the question clients occasionally raise about the material models were built from. A studio is not required to resolve that debate, and it should be able to say which services it used and what those services state about their training and the rights they grant, because a procurement or legal team may ask and an inability to answer reads badly.

What is not settled is worth naming honestly. Jurisdictions differ, guidance is evolving, the threshold of human contribution that suffices has not been tested extensively in commercial contexts, and the position of the services themselves varies and can change. Any studio claiming certainty here is overstating, and any client relying on a supplier's assurance rather than their own legal advice is relying on the wrong thing.

The contractual response is to make ownership explicit rather than assumed. What the client receives, what they may do with it, who owns any 3D assets built for the project, what the tool licences permit, and what happens if a service changes its terms. Sarasvuo et al. (2023) found that buyer perceptions of fit and attractiveness shape how corporate offerings are evaluated in B2B services, and a supplier who addresses this clearly reads as competent rather than as raising problems.

The commercial reality for most projects is that this matters less than clients fear. A corporate film, a product video or an event asset needs a licence to be used, not the ability to prevent others producing something similar. Where it matters is for a distinctive brand asset intended to be defended over years, a character, a signature visual world, a musical identity, and in those cases the answer is to increase the human authorship deliberately and to document it.

The practical position a studio can hold and defend is straightforward. Human creative direction at every stage, recorded. Brand critical elements composited rather than generated. No likenesses of real people without documented consent. Tool licences retained with the project. Full disclosure to the client. Ownership terms stated in the contract. That set of practices produces work with a defensible authorship position and, more usefully, removes the situations in which the unsettled parts of the law would actually be tested.

References

U.S. Copyright Office. (2025a). Copyright and artificial intelligence, Part 2: Copyrightability. https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf

U.S. Copyright Office. (2024). Copyright and artificial intelligence, Part 1: Digital replicas. https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-1-Digital-Replicas-Report.pdf

U.S. Copyright Office. (2025b). Copyright and artificial intelligence, Part 3: Generative AI training (Pre-publication version). https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-3-Generative-AI-Training-Report-Pre-Publication-Version.pdf

Sarasvuo, S., Liljander, V., & Haahtela, K. (2023). Buyer perceptions of corporate brand extension attractiveness and fit in B2B services. Industrial Marketing Management, 115, 69–85. https://doi.org/10.1016/j.indmarman.2023.09.006