In 2023, Scarlett Johansson declined an offer from OpenAI to voice its GPT-4o model. Nine months later, the company launched an AI assistant voice that sounded strikingly like her. Although the company denied using her voice, after facing a public backlash OpenAI ultimately removed the voice option from ChatGPT.
That episode illustrates the new challenges created by generative artificial intelligence (“AI”), which has made it much harder for public figures to control the use and exploitation of their identity. We consulted intellectual property (“IP”) expert and Caravel lawyer, Antoine Malek, to better understand these developments and why people are turning to copyright law as a potential solution.
The deepfake problem
AI tools have made it easy to superimpose a person’s face or voice onto false images or videos that depict them in situations they did not participate in (known as “deepfakes”), without that person’s consent.
The rapid pace of improvement in AI tools has made it easy to create realistic deepfakes that are hard for the public to identify, and then widely disseminate them. Deepfakes can affect anyone, and there is a growing list of examples where they have been used for fraud, harassment, political manipulation, commercial exploitation, or to create non-consensual sexualized images. Public figures are particularly vulnerable to the use of deepfakes because of the commercial and/or reputational value of their identity.
Where the traditional tools fall short
Two bodies of law have traditionally addressed this kind of misuse:
- Right of publicity laws, which protect an individual’s right to control the commercial use of their name, image, likeness, and other aspects of their identity.
- Privacy laws, which can help where they are drafted broadly enough to cover the unauthorized use of a name, likeness, or voice.
However, both right of publicity and privacy laws can vary widely between jurisdictions (e.g. between provinces in Canada, or between states in the U.S.), and thus provide uneven protection. They can also fall short in protecting against the use of a voice or likeness that is merely similar, such as OpenAI’s use of a voice that was similar to Scarlett Johansson’s.
As a result of these shortcomings, public figures are increasingly turning to trademark law to help them control the use of their identity online. Some legislators have also proposed expanding the traditional scope of copyright law to provide a right for individuals to control the use of their identity online.
Public figures are turning to trademark law
Trademark law protects anything that functions as a distinctive commercial identifier. This includes protection for a wide range of non-traditional marks, such as sounds, images, short video clips, or holograms.
Celebrities like Matthew McConaughey and Taylor Swift have relied on this to pursue trademark protection for sound marks consisting of their voices saying famous catchphrases widely associated with them, like Mr. McConaughey’s “alright, alright, alright” or Ms. Swift’s “Hey, it’s Taylor Swift,” or to protect well known images of them. The idea is to prevent others from being able to use their voice or likeness without authorization to sell goods or services that are similar to those for which their trademark has been registered.
This strategy can give public figures nationally enforceable trademark rights to prevent others from using their voice or image, or a very realistic imitation created by AI, to sell products or services. However, the strategy must fit within the legal framework of trademark protection, which is based on the concept of likelihood of confusion and is normally tied to the goods and services covered by a registration.
Trademark law cannot protect one’s likeness in the abstract, the unauthorized use must still be tied to the registered trademark such that it causes consumer confusion about whether the public figure is the source of the goods/services being advertised or sold.
A registered trademark may also be useful in preventing a public figure’s likeness from being used in a non-commercial manner, although this would depend on the specific circumstances. For example, in Canada, a trademark registration provides protection against uses that are likely to have the effect of depreciating the value of the goodwill attaching to the trademark (the concept of “dilution” in the U.S. may similarly be useful, although it protects only “famous” marks). Thus, even in the absence of consumer confusion, a trademark registration may allow a public figure to prevent unauthorized uses of their likeness that nevertheless depreciate the commercial value of their likeness (e.g. by tarnishing their reputation).
As this is a relatively recent and novel use of trademark law, we have yet to see any cases where a public figure uses a trademark registration to combat deepfakes. However, the strategy appears to be gaining traction amongst public figures, with recent trademark filings including the Backstreet Boys’ entertainment company filing a sound-mark trademark application for the spoken words “Hi, we’re the Backstreet Boys,” or Lionel Richie’s trademark applications for spoken words consisting of his famous song lyrics such as “Say You Say Me” or “Easy Like Sunday Morning”.
Denmark copyright proposal
In 2025, Denmark made an unusual proposal to address the problem of deepfakes. The Danish government proposed amendments to its copyright laws to grant individuals the right to control the digital use of their likeness, including their facial features, voice, and other personal characteristics.
Under Denmark’s proposed amendments, enforcement would be governed by the EU Digital Services Act (DSA), which sets EU-wide rules on content moderation that cover detection, flagging, and removal of illegal content to protect users from harm.1
This proposal represents a significant departure from the traditional scope of copyright law, which protects creative works that are original expressions of ideas, such as literary, musical, artistic, and dramatic works. Copyright does not include elements of an individual’s personality or likeness, such as their voice or facial features. The Danish proposal is also inconsistent with the traditional purpose of copyright law, which is to further the public interest by promoting the creation and dissemination of works of the arts and intellect, and to allow creators a just reward for their creations. It is unclear how new rules that protect an individual’s likeness would fit within the framework of copyright exceptions that protect the public interest, and whether such rules could end up unduly restricting the public domain.
So far, the Danish proposal has not been adopted in Europe or in other jurisdictions. This is not surprising given its incompatibility with the legal principles underlying copyright law, and the fact that the risks posed by deepfakes are already addressed to some extent (albeit with gaps) under other legal frameworks, which can include rights of publicity, defamation law, data protection laws, or criminal laws.
A better approach likely involves new legislation designed to specifically address this harm, similar to South Korea’s criminalization of deepfake pornography, or the U.S. Take It Down Act, passed in 2025, which requires digital platforms to comply with notice and removal requirements to ensure that people can quickly and easily get non-consensual intimate images taken down. A commercial equivalent of this “take down” right may well prove more efficient and effective than using IP laws.
Where this leaves us
It remains to be seen how effective IP laws will prove in allowing public figures to combat AI-generated deepfakes and more broadly, control the use of their identities online.
Attempts to rewrite copyright laws for this purpose are misguided and inconsistent with the fundamental purposes of copyright, which are to reward original creative expression and to promote its dissemination to the public. These purposes are incompatible with policing the unauthorized use of aspects of an individual’s identity, such as their face or voice.
Trademark registrations are much more likely to prove useful tools, as they allow for various elements of a person’s identity (such as their voice and image) to be protected, but with the limitation that they can only be protected where they function as commercial source identifiers. However, trademarks are unlikely to replace traditional tools like right of publicity laws or privacy laws, as their fundamental purpose is to protect consumers from confusion in the marketplace and not to protect a person’s persona or identity. Furthermore, their utility is limited to celebrities and other public figures whose personas can be said to have commercial value. Nevertheless, until new legal frameworks are enacted that address the harm of deepfakes directly, their use by public figures as a legal tool to combat deepfakes is likely to grow.
Caravel has a top-tier bench of IP lawyers. Connect with us today to learn more.
Want to stay in the loop? Subscribe to our newsletter for regular legal and business updates.