Addressing the limitations of the right of publicity in the world of AI

A couple of weeks ago, actor Tom Hanks posted a screenshot of an AI-generated deep fake video of himself with the caption “BEWARE!! There’s a video out there promoting some dental plan with an AI version of me. I have nothing to do with it.”

That a person with the resources and representatives of Tom Hanks would address a commercial misappropriation of his likeness via a social media post highlights a real disconnect between the law and technology that is fast-outpacing it. Sure, Mr. Hanks had a slam dunk claim under California’s right of publicity statute (Cal. Civ. Code Section 3344), which provides for damages, disgorgement of profits, punitive damages and attorneys’ fees where a person “knowingly uses another’s name, voice, signature, photograph, or likeness, in any manner,” for commercial purposes, without consent. Indeed, at the moment, the right of publicity is virtually the only enforcement tool available to actors and musicians whose names, likenesses, or voices are recreated or mimicked by AI without their consent. However, it is not a particularly effective tool.

At present, the right of publicity is a state-by-state right that is either statutory or part of a state’s common law. There is no federal or international right of publicity. Because of this, the scope of the right and the remedies available vary drastically from state to state. For example, some states, such as California, recognize posthumous rights of publicity while others either do not (or differ in terms of their duration). It is also highly likely that many of the “bad actors” who are creating deep-fake videos of celebrities are located overseas, which would mean that a person such as Hanks would have to worry not only about establishing jurisdiction in California but also evaluate the likelihood of being able to enforce any injunctions or judgments that might be entered by a California court.

Even if you get past those hurdles, filing a lawsuit and obtaining interim and final relief takes time. This past spring, a deepfake song using AI-generated vocals intended to mimic Drake and the Weeknd was released online. The creator of the song had used original lyrics but included underlying music that was copyrighted by a third party. Because of that, the copyright holder’s record label was able to make a Digital Millennium Copyright Act (DMCA) claim and get the song taken down. Before it was removed, however, it was played hundreds of thousands of times on Spotify and many millions of times on TikTok. Had there not been a copyright hook for a DMCA claim, the musicians’ remedy would have been via Civil Code 3344, or a sound-a-like case in the vein of Midler v. Ford Motor Co., 849 F.2d 460 (9th Cir. 1988) (Midler prevailed on a theory that she had a common law right in the identity of her voice which was deliberately imitated without her consent). While such a claim would certainly be successful, the time it would take to get it filed, obtain a TRO and use that TRO to have the song removed from all platforms would take infinitely longer than a simple DMCA takedown (and many more millions of streams would occur in the interim).

At the moment, it is often easier to set the record straight via Instagram than to actually pursue a legal claim. That might work for Hanks, whose fans will not be duped into buying a dental plan once they know he is not behind it. For a musician, however, it’s less clear, as there are many people who might just decide they like the song, whether or not it was authorized and decide they want to listen to it anyway. Because of the way that streaming services compensate artists, there is a real concern that a flood of AI-generated content will reduce revenue to human artists and, ultimately, yield a situation of less human creation and innovation. The music industry has historically been at the forefront of legal disputes and the evolution of law with changing technology. The question becomes, what can be done this time?

For one thing, musicians and actors should be negotiating contracts that place limits on what studios and record labels can do with video and audio captured of them in connection with their performance of services on a particular work. If they have not done so already, platforms that host music or other content should consider updating their terms and conditions to make it a violation to post/stream content that is not wholly (or at least substantially) created by a human.

It is also time to seriously consider the legislation of a federal right of publicity, which would provide for uniformity of the law and also establish federal question jurisdiction allowing for plaintiffs’ use of the federal courts (at present, you can only bring a right of publicity claim in federal court if there is diversity jurisdiction or a basis for supplemental jurisdiction).

Finally, we should be passing legislation to create something akin to the DMCA for deepfake images, videos, and songs that are presently not subject to DMCA takedown because they do not legally constitute copyright infringement. It should be the case that a person such as Hanks could make a quick, simple online report based on his ownership of an interest in his likeness and voice that triggers the immediate and automatic take-down of the content at issue while its authenticity is assessed in the background. The same notice and counter-notice procedures used for DMCA takedowns can be used to assess right of publicity claims. The DMCA works fairly well as a mechanism for quickly halting copyright infringement online. An expansion to cover unauthorized uses of names, likenesses, and voices will be essential to combat the proliferation of AI-generated infringements, particularly when you do not know where they are coming from.

While we cannot fully anticipate how the technology will evolve, it is critical that we take these steps now so that we can tell our clients that there is an option other than the drastic (and slow) remedy of filing a lawsuit, or posting something on social media that draws more attention to the fact that the client’s likeness or voice was misappropriated. Ironically, the real Hanks created more press for that dental plan than the AI Hanks ever did.