This article was originally published by WTR linked here.
Discover the six-step clearance framework to help brands assess when AI-generated identities become a legal and reputational liability.
Picture this. A global brand launches a new campaign fronted by a striking ambassador. She looks real, speaks naturally and recommends the product in the first person. But she does not exist. Her face, voice and personality are entirely AI-generated.
At first, the legal advantages seem obvious: no talent negotiations, no scheduling conflicts, no diva moments. But the simplicity is an illusion. What if her face resembles a well-known actress? What if her voice sounds like a recognisable celebrity? What if consumers believe she is a real person sharing a genuine experience? And what happens when the campaign crosses borders?
These scenarios illustrate why AI-generated identity cannot be treated as just another advertising asset. It sits at the intersection of legal interests that brand teams have traditionally reviewed separately: identity and likeness, trademark and endorsement, copyright, and consumer protection. But AI adds a variable that none of those frameworks were originally designed for: what does the consumer believe is real?
The question for brands is no longer just whether they have the right to use a particular face or voice. It is who gets to create, borrow and commercially exploit the trust that an identity generates.
The AI Identity Stack
A synthetic identity is not a single asset that is either ‘cleared’ or ‘not cleared’. I find it more useful to think of it as a stack, four overlapping layers, each of which needs attention before an AI-generated identity goes live.
First, identity
Does the persona reproduce or evoke a real person? The test is not always whether an exact replica was created. A synthetic face paired with a recognisable voice, set of mannerisms or distinctive style can trigger an association even though no photograph or recording was ever copied.
Second, brand and endorsement
Could consumers perceive an affiliation or sponsorship that does not exist? This takes the analysis beyond likeness rights into trademark, false endorsement and unfair competition territory.
Third, content.
What sits behind the final execution? Images, recordings, music and other protected material used to build the asset may each require their own copyright and licensing clearance.
Fourth, consumer trust
Could consumers believe the spokesperson is real? That a testimonial reflects an actual experience? That a recognisable person authorised the endorsement? Or that an AI-generated depiction demonstrates a result the product can actually deliver? These layers are cumulative. As opposed to most trademark work, clearing one does not clear the others.
Same campaign, different rules
The stack is a risk-flagging tool, not a universal legal test. The answer to each question depends on where the content runs.
Take identity. A synthetic persona that evokes a real individual may be analysed through publicity rights in one jurisdiction and through image, personality, privacy or constitutional rights in another. The terminology, scope and remedies will differ.
Transparency adds another dimension. The EU AI Act, for example, now imposes disclosure obligations for certain AI-generated or manipulated content, including deepfakes. The significance for global brands is not just that Europe has set a standard. It is that the use of AI can itself become information that shapes how an audience understands content.
Latin America is arriving at similar questions by a different route. Most jurisdictions in the region do not yet have an EU-style horizontal AI regime. But the absence of AI-specific legislation does not mean advertisers can afford to wait.
Consumer and advertising laws across the region are generally technology-neutral. Existing requirements that advertising be truthful, clear and non-misleading can address AI-generated content without ever mentioning “artificial intelligence”. This matters most when AI changes something material: whether the person depicted exists, whether a testimonial is real, whether an endorsement actually occurred, or whether a product result has been artificially created.
As AI regulation develops in the region, EU transparency rules may provide a reference point. But existing advertising and consumer protection principles already cover much of the same ground. The practical lesson is that brands should not wait for AI-specific labelling rules before adopting disclosure practices. Where AI materially shapes what consumers see or hear, existing advertising and consumer protection frameworks may already require transparency. The absence of a dedicated AI labelling obligation does not mean the obligation does not exist.
Not every use of AI is equal
This distinction matters more every day as AI becomes embedded in ordinary creative tools. Brands already use AI to translate copy, resize images, remove backgrounds and adjust lighting. Slapping a label on every advertisement that has been ‘touched by AI’ would tell consumers nothing useful.
The real question is whether AI created or materially altered something relevant to what the consumer takes away from the communication. Removing a shadow is not the same as generating a model who does not exist. Translating an approved statement is not the same as making a real spokesperson appear to say words she never said. Correcting an image is not the same as fabricating a visual demonstration of product performance.
The distinction between using AI and materially using AI should be part of every advertising clearance.
The six-step clearance roadmap
For brand and trademark teams, the stack translates into a six-stage process: identify; trace; assess; clear; disclose; and document. Think of it as the AI cousin of the trademark clearance you already know. Same discipline, different materials.
Identify
What kind of identity is being used: a real individual, a licensed digital replica, a fully synthetic persona or a hybrid? Does it evoke someone identifiable through appearance, voice, name or mannerisms?
Trace
Understand where the asset came from. What inputs, reference materials or existing content were used? “AI-generated” should never become shorthand for “origin unknown”.
Assess
Review the finished communication through each layer of the stack. Is an individual’s identity implicated? Could endorsement be inferred? Is protected content involved? Has AI created or materially altered something the consumer may rely on?
Clear
Determine which permissions are required under applicable local law. Where real people are involved, check whether existing releases adequately address AI-enabled modification, digital replicas, new media, territory and duration.
Disclose
Do not just ask whether AI was used. Ask whether consumers need to know how it was used to avoid a misleading impression. A synthetic identity or artificial testimonial demands a very different analysis from routine AI-assisted production.
Document
Preserve context, permissions, territorial restrictions and disclosure decisions. AI-generated assets can be modified, localised and redeployed almost instantly. Clearance for one execution should not silently become clearance for every future iteration.
Protecting trust, not just the avatar
Generative AI gives brands an unprecedented ability to create and manipulate identity. It does not eliminate the legal interests that attach to identity, endorsement, content and consumer perception.
The regulatory answer will not be identical everywhere. Some jurisdictions will address these questions through rights over the individual, others through trademark and consumer protection, and some are already imposing AI-specific transparency requirements.
For global brand teams, the goal should not be to find a single rule. It should be to build a clearance process that spots the issue, identifies when AI matters to the communication and asks the right question under the applicable local law. The question is no longer just who owns the face, voice or avatar. It is who gets to create, borrow and commercially exploit the trust that identity generates, and when consumers need to know that what they are trusting is artificial.
