Generative artificial intelligence has made it easier than ever to replicate someone’s voice or face without their consent, and the law has struggled to keep pace. In response, a bipartisan group of senators introduced a revised version of the Nurture Originals, Foster Art, and Keep Entertainment Safe (NO FAKES) Act on May 20, 2026, marking a federal legislative effort to protect individuals’ voices and visual likenesses from unauthorized AI-generated digital replicas, or “deepfakes.” For AI platforms, content distributors, and companies operating across digital media, entertainment, consumer technology, and advertising, this proposed legislation may come with new compliance obligations while also adding some certainty missing from the current patchwork system.
What Would the Revised NO FAKES Act Do?
The revised NO FAKES Act would create a federal liability framework targeting the production and distribution of nonconsensual AI digital replicas. Under the bill, individuals and companies would face potential liability for distributing an unauthorized digital replica of a person’s voice or visual likeness in an audiovisual work or sound recording. It would also create a framework for on-platform takedowns.
Compared to the bill’s prior version, the revised legislation adds three meaningful updates:
- It introduces a counter-notice procedure designed to safeguard free speech rights, providing a mechanism for those who believe their content was wrongly removed to seek reinstatement.
- It carves out an exemption for libraries, archives, and research institutions, ensuring that the study and analysis of digital replicas is not inadvertently chilled.
- It addresses technical gaps affecting streaming music platforms, bringing greater clarity and certainty to how the bill would operate in that context.
A Layered Identity-Protection Landscape
The NO FAKES Act would largely preempt the existing patchwork of existing legal frameworks that brands, talent, and their counsel are currently maneuvering.
The most natural legal vehicle for protecting an individual’s voice and likeness is the right of publicity, the legal right to control the commercial use of one’s own identity. But right-of-publicity protections are state-based and some states, like New York and California, have robust statutes, while others don’t. Copyright law is similarly limited in this context. Copyright protects original works of authorship, not the underlying human attributes, i.e., a person’s face, voice, and mannerisms.
These limitations have sent some high-profile talent exploring trademark law as a supplemental shield. Matthew McConaughey made headlines when he secured federal trademark registrations for elements of his voice and likeness, and Taylor Swift has followed with similar filings, both operating on the theory that federal trademark rights could provide a broader and more uniform deterrent against AI-generated replicas than state-level right-of-publicity claims alone. Even so, trademark law may not provide sufficient protection, since it was not designed to address the composition of AI-generated content directly. The NO FAKES Act would close that gap by creating a federal private right of action specifically targeting unauthorized AI-generated digital replicas.
Finally, several states have enacted laws with direct relevance to AI-generated digital replicas. Tennessee’s ELVIS Act, in effect since July 2024, protects voice likenesses from unauthorized AI replication and applies broadly to any individual, not just musicians. California’s AB 2602 addresses AI replication of performers in entertainment contracts, and California’s existing right-of-publicity statute has been interpreted expansively in the AI context. New York’s right-of-publicity law similarly covers AI-generated uses. These protections have been the primary tool for entertainers and public figures fighting misappropriation, but they apply only within individual states and produce inconsistent results.
Companies in affected industries may want to concentrate their initial compliance efforts on the areas where the NO FAKES Act’s proposed standards align with what these state laws already require. Identifying these areas and building a corresponding compliance infrastructure would satisfy current legal obligations and position the company for the federal regime that may follow. Practical priorities in this phase could include reviewing which state laws apply to current operations, auditing vendor and platform agreements for digital replica provisions, and confirming that consent and authorization workflows cover the uses those statutes regulate.
Practical Takeaways
Though it has a ways to go, the bill is gaining visibility and bipartisan support. The next meaningful signal will be whether the Senate Judiciary Committee schedules hearings or a markup.
Regardless of the bill’s next steps, brands and talent should consider creating an inventory of commercial identity assets, including licensed voice recordings, authorized likeness approvals, and existing contractual permissions covering digital or synthetic uses. If those agreements predate the generative AI era, they likely need updating. Establishing a clear record of authorized uses now may help to assert rights and demonstrate compliance if the bill becomes law.
Likewise, for AI platforms and content distributors, the focus should be on assessing platform liability exposure. The bill’s knowledge-based liability standard for hosting unauthorized replicas would require robust content identification, takedown, and counter-notice processes. Platforms familiar with Digital Millennium Copyright Act safe harbor frameworks will recognize the structure, which we may see platforms implement as a protective measure even if the bill does not pass.
In-house legal and compliance teams should consider how a federal NO FAKES Act cause of action would interact with existing right-of-publicity, trademark, and copyright protections. A claim under the NO FAKES Act could run alongside existing IP and publicity claims, not in place of them, and risk frameworks should reflect that layered reality.