It sounds like science fiction, but it is real. TikTok creator Khaby Lame recently agreed a US$975 million deal with a Hong Kong investment business. The reason? It wants to use his face, not only for traditional advertising but also to create an AI version of him. Welcome to an era in which your likeness has become a commodity.
From Nespresso advertisements to AI avatars
The idea that a famous face generates revenue is not new. George Clooney has been drinking Nespresso on screen for years. What is new is that businesses no longer want merely to film your face. They want to recreate it, copy it and make it speak languages you do not speak yourself, 24 hours a day, without your presence.
Khaby Lame, the Italian TikTok creator who became famous for remaining silent in his videos, will now speak through AI—in several languages—and livestream without sitting in front of a camera himself. Hong Kong company Rich Sparkle expects to earn more than US$4 billion from him over three years. The deal means it decides where his face appears and what he participates in.
Bruce Willis as a digital twin
Khaby Lame is not the first. American actor Bruce Willis participated in a similar arrangement in 2022. A Russian telecoms company used 34,000 photographs to create a digital copy of his face. That digital twin then appeared in advertising campaigns without Willis needing to be on set.
The trend is visible closer to home too. Former footballer Ruud Gullit recently launched an AI clone of himself specifically for brand collaborations. At the launch, he said this would be entirely normal in five to ten years. He is probably right.
The sporting world is joining in
The trend extends beyond individual celebrity deals. In the United States, the Major League Baseball Players Association, representing professional baseball players, has partnered with technology company Genies to create AI characters of leading baseball stars.
Fans will be able to interact with digital versions of their favourite players throughout the year, through numerous digital channels as well as around games. This is described as a shift from passive viewing to active interaction. The players need do nothing themselves: their digital counterparts take over.
The quiet army of AI models
Alongside the famous names, another group is participating in this development. Professional models and actors sell their faces, expressions and body language to platforms such as Synthesia and Hour One. Their faces become AI avatars appearing in corporate videos, e-learning and advertisements worldwide.
These models often agree a one-off buyout. Afterwards, they have little or no say over how their digital likeness is used. An attractive short-term payment may create difficulties over the longer term.
The legal framework: what are portrait rights?
To understand these deals legally, we need to return to the basics. In the Netherlands, Article 21 of the Copyright Act governs portrait rights. It protects against publication of your portrait without permission where you have a reasonable interest in opposing it. In commercial use, such an interest will, broadly speaking, almost always exist.
There is more to the legal position. An AI clone often has a dual legal character: both a portrait and personal data. Once a digital version relates to an identifiable person, the GDPR also applies. A valid processing basis, usually consent, is then required.
A portrait is broader than you might think. Recognisability is the key. A photograph, video, illustration, 3D model or even a character in a series may constitute a portrait. The question is always whether the person is recognisable.
Interestingly, case law does not treat a voice as the subject of portrait rights, although it may constitute special-category personal data under the GDPR. An imitated voice therefore falls under a different protective regime.
The distinction between commissioned and non-commissioned portraits also matters. The first category involves contractual arrangements. In the second, interests must be balanced: the portrayed person’s privacy, honour and reputation on one side, and the interest in publication on the other. That balance determines whether publication is permitted.
What about AI-generated faces?
AI can generate new faces from a large collection of photographs. Technically, those faces are unique. But what happens if a generated face closely resembles an existing person?
In 2022, the Dutch Supreme Court delivered an important judgment in the dispute between Max Verstappen and delivery service Picnic. Picnic had used a Verstappen lookalike in an advertisement without his permission. The Supreme Court held that using a lookalike may also constitute use of someone’s portrait.
That judgment is relevant to AI-generated faces. If an AI portrait closely resembles an existing person and is used commercially without permission, portrait rights may be infringed. The face need not be identical. Sufficient resemblance for the public to recognise the person may be enough.
The darker side: deepfakes without consent
Deepfakes illustrate the darker side of the technology. A pornographic deepfake of television presenter Welmoed Sijtsma was created without her consent. Her face was placed on the body of an adult-film performer, and the video spread through websites and WhatsApp groups.
Sijtsma reported the matter to the police and made the NPO 3 series Welmoed en de sexfakes, in which she searched for the creator. With an IT investigator’s help, she identified the perpetrator. The court described the video as shocking and tasteless and found that the victim’s sexual autonomy and privacy had been seriously harmed.
The defendant’s lawyer argued that creating deepfakes had not expressly been criminalised. The court rejected that argument. Deepfakes can also fall within the criminal prohibition where the material appears sufficiently real that its manipulation is not immediately apparent. The creator ultimately received a suspended community-service sentence of 120 hours.
Alongside criminal law, the portrayed person may rely on portrait rights. Options include prohibiting dissemination and claiming damages. This illustrates an important point: not every AI portrait is problematic, but the position changes when it casts someone in a false light or is used commercially without permission.
A legislative proposal would introduce a related right protecting against deepfakes. If adopted, it would give each person the right to prohibit or authorise creating and disseminating deepfakes of their voice or appearance.
Virtual influencers and deception
Deepfakes recreate existing people. There is also another side: wholly fictional people who have never existed. Earlier this year, an Instagram account presenting a supposed American soldier, Jessica Foster, grew to more than a million followers within months. She posed in barracks, beside fighter aircraft and apparently even alongside world leaders. Everything looked real, but none of it was. The account proved to be entirely AI-generated and was used to direct followers to paid platforms offering explicit content.
This kind of deception occupies a legal grey area. There is no existing person whose portrait rights are infringed. Consumers may nevertheless suffer harm, potentially involving unfair commercial practices or fraud. It also demonstrates how convincing generative AI has become. Small visual errors, once a useful clue, are increasingly rare. Social media is entering a new phase: entire identities, as well as individual posts, can be fabricated.
Can you actually sell your portrait rights?
Returning to the deals, headlines about selling your face are somewhat misleading. Although portrait rights appear in the Dutch Copyright Act, they are not intellectual property rights. Your portrait right is not property that can simply be transferred.
What Khaby Lame and others are actually doing is granting an extensive licence. They permit a business to use their face, often exclusively, often for a fixed period and increasingly with a right to create AI versions. The portrait right itself nevertheless remains with them.
This sounds lucrative, and it is. But there is an important downside. You are present at a traditional photo shoot or commercial, can see what happens and can intervene. AI-generated content is different. A business can rapidly create new images without your input. Your face may appear in contexts you would never have chosen.
Imagine your AI avatar promoting a product you have no affinity with. Or worse, your digital twin saying things directly contrary to your beliefs. These scenarios require contractual safeguards because intervention afterwards is difficult. New techniques make certain things easier for the business but more complicated for the person whose face is used.
Who is responsible for what an AI clone says?
An often overlooked question is who bears responsibility if your AI version says something unacceptable. In practice, primary responsibility usually lies with the party deploying the clone, rather than the person on whom it is based. Under traditional liability law, the business developing, training and exploiting the system is also the first party to face a claim. This applies even more strongly if the clone learns independently and the portrayed person has no direct control over its statements. Nevertheless, address this expressly in the contract: allocate responsibility for training, supervision and publication, and include an indemnity for potential loss.
Key points for AI portrait licences
Anyone considering such a deal should address several crucial points.
First, the licence’s scope. What exactly may your face be used for—and, equally importantly, what is prohibited? The more specific, the better. Consider geographical limits and types of content too.
Second, duration. Three years of exclusivity sounds manageable, but what happens afterwards? May the business continue using previously created AI content, or must everything be taken offline?
Third, control. Can you review the final product before publication? Do you have a veto? These arrangements can distinguish a successful collaboration from a reputational nightmare.
Fourth, remuneration. A one-off buyout seems simple, but if your avatar generates income for years, you may want to participate. Consider royalties or profit sharing, or, as with Khaby Lame, an equity stake enabling you to share in the success.
Fifth, protection against misuse. What if your AI portrait is used in a way that harms your reputation? What legal remedies are available, and who bears liability?
Finally, what happens after the relationship ends? Can the business still use the clone, for how long, and is additional payment due? Answer these questions beforehand, not afterwards.
A market still in its infancy
This development is only beginning. Technology is becoming better and cheaper, making more than A-list celebrities attractive for such deals. Micro-influencers, presenters, athletes and models are being approached too.
If you are considering an arrangement, understand what you are agreeing to. Taking your face back from the internet is considerably harder than making it available.
