When an Athlete’s Image Is No Longer Their Own

For decades, public figures have made money from something that is simultaneously valuable and impossible to separate from them: their identity. Athletes sign endorsement agreements allowing companies to use their names, images, voices, and likenesses. Actors license their appearances, musicians commercialize their voices, and influencers build businesses around personalities that audiences immediately recognize. Artificial intelligence is complicating the basic assumption underlying all of those arrangements.

Historically, using someone's identity commercially required some form of access to that person. A company wanting an athlete in an advertisement generally needed the athlete to participate, provide photographs or recordings, or authorize existing material. Even unauthorized uses usually depended on copying something the person had actually created. Generative AI removes much of that constraint. An athlete's face can now appear in a video they never filmed, their voice can say words they never spoke, and their likeness can endorse a product they have never used or express an opinion they do not hold. None of it requires the athlete to participate.

This is not a problem unique to athletes. Any sufficiently recognizable public-facing person can be affected, including actors, musicians, politicians, journalists, executives, and online creators. The regulatory questions surrounding consent, identity, fraud, and synthetic media therefore extend far beyond sports. Athletes, however, provide a particularly useful case study because professional sports already have sophisticated systems for assigning economic value to identity. An athlete's name, image, likeness, and reputation can be separately licensed, negotiated, protected, and monetized. AI is entering an environment in which the underlying asset already has a price.

Recent developments have made the problem increasingly difficult to treat as hypothetical. During the 2026 World Cup, monitoring of online advertising identified thousands of suspected deepfake advertisements using the identities of prominent football players, including synthetic content appearing to show players promoting gambling and investment products. The athletes depicted had not made those endorsements.

The most obviously fraudulent examples are relatively straightforward from a regulatory perspective. Someone should not be able to fabricate an athlete endorsing a financial product and use that fabrication to deceive consumers. The harder questions begin when the use becomes less obviously fraudulent. A fan might create a convincing AI video of an athlete for entertainment. A brand could produce an AI-generated version of an athlete while clearly disclosing that it is synthetic. A broadcaster might recreate a player's voice to provide content in another language, while a video game could generate realistic athlete interactions that were never individually recorded.

Some athletes may actively want those opportunities. An athlete could license an AI version of herself to communicate with fans in multiple languages, participate in advertising without attending every production shoot, or appear in interactive experiences that would be impossible to record individually. Retired athletes could similarly authorize synthetic voices or appearances for documentaries, games, or other commercial projects.

AI therefore creates both an identity problem and an identity market. The regulatory objective should not simply be to prevent synthetic versions of people from existing. It should be to preserve a meaningful distinction between authorized replication and unauthorized appropriation.

Professional sports are already beginning to experiment with the authorized side of that market. Player associations and technology companies have explored AI-powered avatars, digital replicas, and other systems that allow athletes to license synthetic versions of themselves. These arrangements suggest that digital identity could eventually become another commercial right negotiated alongside traditional endorsements, licensing agreements, and media appearances.

The problem is that generative AI also changes the economics of unauthorized use. Traditional endorsement disputes generally involved identifiable transactions: a company used an athlete's photograph without permission, or an advertisement falsely suggested an endorsement. The athlete or their representatives could identify the use and potentially pursue whoever was responsible.

Synthetic content can instead be created cheaply, distributed globally, altered almost instantly, and reproduced by enormous numbers of users. One athlete could theoretically appear in thousands of fabricated advertisements, videos, voice recordings, memes, and promotional materials without knowing that most of them exist. An enforcement system that depends entirely on the athlete discovering every unauthorized use and pursuing it individually becomes increasingly unrealistic.

That shifts part of the regulatory question toward platforms and AI providers. If a platform receives credible notice that a video is impersonating an athlete to promote a fraudulent investment product, how quickly should it be expected to act? Should AI systems preserve technical information that allows synthetic content to be identified? Should commercial uses face stronger disclosure and consent requirements than ordinary users creating obvious parody or entertainment?

The European Union has begun addressing part of this problem through the AI Act. Transparency requirements applicable in 2026 require certain artificially generated or manipulated content to be identifiable as such and establish disclosure obligations for deepfakes, with modified treatment for some artistic, fictional, satirical, and similar content. These requirements address an increasingly important problem: audiences need some way of knowing when realistic content has been artificially created or manipulated.

Disclosure, however, does not resolve the underlying question of consent. An advertisement could accurately identify itself as AI-generated while still using an athlete's likeness without permission. A label tells the audience something about how the content was created; it does not necessarily establish that the person being depicted agreed to appear in it. Transparency and consent are related regulatory objectives, but they are not interchangeable.

That distinction is particularly important in sports because endorsement value frequently depends on exclusivity. Imagine an athlete has a multimillion-dollar agreement with one sportswear company and an AI-generated advertisement depicts that athlete praising a competitor's shoes. Even if the advertisement is labeled as synthetic, it derives its commercial value from the audience's recognition of that particular athlete. The athlete's likeness has value precisely because consumers associate the face, voice, and reputation with the real person. AI does not eliminate that association; it relies on it.

Synthetic voice presents an even more complicated problem because audio can circulate without the visual cues that sometimes cause audiences to question manipulated video. A short recording appearing to capture an athlete criticizing a coach, discussing an injury, requesting a trade, or making an inflammatory statement could spread widely before its authenticity is established.

In sports, those consequences can extend beyond reputation. Information about injuries, trades, availability, and team relationships can influence betting activity, fantasy decisions, ticket demand, and perceptions of both players and organizations. A convincing synthetic recording suggesting that a quarterback is injured or a basketball player intends to request a trade could therefore have economic consequences before anyone establishes that the recording is false. Synthetic identity can become a market-integrity problem as well as a personal one.

Sports organizations may eventually need to think beyond removing deepfakes after they appear. Authentication could become just as important as enforcement. Leagues already maintain official websites, social-media accounts, injury reports, transaction systems, and communications departments. As synthetic media becomes more convincing, establishing reliable ways to verify authentic athlete communications could become another part of that infrastructure.

That might involve mechanisms through which official content can be authenticated, rapid-response procedures for disputed media, or agreements with major platforms for handling credible impersonation involving athletes and league personnel. None of this should make leagues arbiters of everything said about their players. Parody, criticism, commentary, satire, and fan-created content are ordinary parts of sports culture and raise important expressive interests of their own.

The more useful distinction is between expression and impersonation. A clearly absurd AI video showing a basketball player competing on the moon is unlikely to convince viewers that the event actually happened. A realistic video showing the same player apparently endorsing an investment platform presents a fundamentally different problem. Regulation that treats those uses identically risks becoming either ineffective against harmful impersonation or unnecessarily restrictive toward legitimate expression.

Consent provides an especially useful starting point for commercial uses. If an athlete authorizes a digital replica, the agreement should establish what that replica is permitted to do. Permission to create an AI avatar for a video game should not automatically become permission to use the same likeness in advertising. Permission to synthesize a voice in another language should not necessarily allow that voice to discuss political issues, endorse unrelated products, or appear in contexts the athlete never contemplated.

This is where the next generation of athlete contracts may become considerably more complicated. Traditional name, image, and likeness agreements primarily address where, when, and how a person's identity can be used. AI agreements may also need to determine what a synthetic version of that person is permitted to say, do, endorse, and become. Athletes may need contractual rights to approve certain uses, prohibit particular subjects, revoke authorization, audit how their digital replicas are being deployed, and determine whether synthetic versions of themselves can continue operating after a commercial relationship ends.

That final issue illustrates how different AI replication is from traditional advertising. A photoshoot produces a finite collection of material, and a commercial is filmed at a particular moment. When an endorsement agreement expires, the company's ability to create entirely new performances by the athlete would traditionally end with it. A sufficiently sophisticated digital replica changes that assumption because it may remain capable of producing new material long after the original relationship has ended.

This creates questions that ordinary image-rights agreements were never designed to answer. If a company trained or developed a model capable of reproducing an athlete's face or voice during a licensing agreement, must that model be deleted when the agreement expires? Can it be retained but not used? Can the underlying technology be transferred to another company? If the athlete later signs an exclusive agreement with a competitor, what happens to the previously created replica?

The same questions become even more complicated after an athlete retires or dies. Sports already commercialize history extensively through video games, documentaries, archival broadcasts, advertisements, collectibles, and anniversary campaigns. AI could make those representations interactive and effectively unlimited. A retired or deceased athlete could theoretically appear in entirely new advertisements, deliver synthetic interviews, or interact with fans decades after their playing career ended.

Existing publicity and intellectual-property laws provide some protection, and some jurisdictions recognize commercial rights that continue after death. But generative AI changes both the realism and scale of what can be created. The regulatory question is no longer limited to whether someone can reproduce an existing photograph or recording. It increasingly concerns whether someone can create an entirely new performance by a person who never performed it.

Again, athletes are not unique. Actors face synthetic performances, musicians face voice cloning, journalists and politicians can be made to deliver fabricated statements, executives can be impersonated in financial scams, and online creators can appear to endorse products they have never encountered. Society built many of its existing rules around a world in which convincingly reproducing another person's identity was expensive, technically difficult, and relatively easy to trace. Generative AI is changing those assumptions.

Sports may be one of the first industries forced to confront the consequences because athlete identity is already so extensively commercialized. Leagues, unions, sponsors, broadcasters, gaming companies, and athletes routinely negotiate over who can use a player's identity and how much that permission is worth. AI does not make those rights obsolete. It makes defining their boundaries more urgent.

The regulatory response should therefore avoid two extremes. Treating every synthetic depiction as inherently wrongful would ignore legitimate uses ranging from satire to athlete-authorized commercial products. Treating disclosure as sufficient would ignore the fact that someone's identity can still be commercially exploited even when everyone knows the resulting content is artificial.

A more durable framework would distinguish between transparency, consent, and harm. Audiences should be able to know when realistic content has been artificially generated. Individuals should retain meaningful control over commercial uses of their identities. Platforms should have workable procedures for responding to deceptive impersonation, while contracts authorizing digital replicas should define the boundaries of that authorization rather than granting indefinite control over a person's synthetic identity.

For athletes, collective bargaining and player associations may become particularly important. Individual athletes negotiating with sophisticated technology companies may not always anticipate how valuable a digital replica could become or how many future uses might emerge from a seemingly narrow agreement. Collective standards could establish baseline protections involving commercial replication, information about how digital models are being used, procedures for withdrawing authorization, and restrictions on transferring synthetic identities to third parties.

Sports organizations have their own reason to take the problem seriously. An athlete's reputation contributes to the value of the league itself. Fans follow recognizable personalities, sponsors pay for association with trusted athletes, and broadcasters build narratives around them. If audiences become unable to distinguish authentic athlete communications from synthetic ones, the damage does not stop with the individual whose likeness was copied. Trust is part of the product professional sports sell.

That is why synthetic identity should not be treated solely as a personal privacy problem or a new category of intellectual-property dispute. It is also a governance issue. The institutions that profit from athlete identity have an interest in protecting the conditions that make that identity valuable.

Athletes have spent decades negotiating greater control over their names, images, likenesses, and commercial opportunities. Generative AI introduces a strange new possibility: the most commercially active version of an athlete may eventually be one that does not require the athlete to be present at all. That could create enormous opportunities when the athlete chooses it, but it creates a very different problem when someone else makes that choice for them.

The central regulatory challenge is therefore not whether AI should ever be allowed to reproduce human identity. In many circumstances, people will actively want it. The challenge is preserving the person's authority over what their synthetic identity is allowed to become.

For athletes, the next generation of image rights cannot stop at controlling where their photograph appears or which company can put their face on a billboard. Those rights increasingly need to address something far more expansive: who can reproduce them, what that reproduction can do, and how long it is allowed to exist.

In the age of generative AI, protecting an athlete's image may no longer mean protecting a picture. It may mean protecting the athlete's ability to remain the person who decides what their own face and voice are allowed to say.

*Photo courtesy of The University of Rhode Island