Skip to content
Post

AI turns images into adverts: what companies are and aren’t allowed to do

Competition law Data protection law Intellectual property Advertising & Sales Press & Media

An existing video can be transformed into a new advert in a matter of minutes using AI. However, just because the source material is publicly available does not mean it can be used for advertising, altered or accompanied by new statements. Anyone who uses AI to place recognisable individuals in a commercial context must pay particular attention to personality rights, data protection, copyright and the rules against misleading advertising. Otherwise, a creative campaign can quickly turn into a legal and financial risk.

Publicly available does not mean it can be used freely for advertising

Videos, interviews, photos and clips from publicly accessible media are easy to find these days and just as easy to manipulate using technology. Generative AI makes it possible to alter faces, adjust lip movements, replicate voices, place people in new situations or create entirely new advertising content from existing video material.

However, technical feasibility says nothing about legal permissibility. Particularly in the case of image and video material, several distinct rights must be taken into account. This concerns, on the one hand, the person depicted or speaking and, on the other, those who hold rights to the original photograph, video or audio material.

The fact that a person has voluntarily made themselves available for a public interview, an event or another form of publication does not, therefore, automatically mean that they have also consented to the subsequent commercial use of their image, voice or statements. Nor can consent to the material being altered using AI and placed in a new advertising context be readily inferred from this.

In fact, the Fair Trading Commission has also explicitly stated in Principle No. B.7, paragraphs 1 and 4, that it is unfair to use personal data such as a name, likeness, statement or voice in commercial communications without express consent. Furthermore, according to the Swiss Fair Trading Commission, failure to comply with data protection rules is also unfair.

The right to one’s own image also applies in the age of AI

The right to one’s own image forms part of Swiss personality rights. The Federal Data Protection and Information Commissioner (FDPIC) expressly states that, in principle, every person may decide whether and in what form their image may be recorded and published. Furthermore, as soon as a person is recognisable in a photograph, this constitutes personal data. The right to one’s own image described above therefore also extends to a person’s voice and spoken words in the sense of an independent ‘right to one’s own voice’ as a component of the right to privacy under Article 28 of the Swiss Civil Code (ZGB).

Under Article 28 of the Swiss Civil Code (ZGB), an infringement of personality rights is unlawful unless it is justified by the consent of the person concerned, by an overriding private or public interest, or by law. The Federal Supreme Court also treats the right to self-determination regarding the publication of one’s own image – and in particular its use in commercial advertising – as an integral part of the protection of personality rights.

For businesses, this means: If a recognisable person is to be specifically featured as part of an advertising campaign, obtaining clear consent is generally the most legally secure approach. The FDPIC expressly recommends written consent for commercial uses, specifying the manner in which the images are to be published.

Existing consent does not automatically cover AI advertising

Particular attention should be paid to the scope of any existing consent. For example, someone who has consented to the publication of an interview, a company profile or an event video has not necessarily also authorised its use for paid advertising.

The decisive factor is what the data subject was informed about and the specific context of use to which they consented. Valid consent requires adequate information, with the context of publication playing a particularly central role.

For new productions, therefore, the consent should specify as precisely as possible whether the material may be used for advertising and whether editing using artificial intelligence is permitted. Depending on the campaign, it should also be clarified whether the person’s face, voice, movements or statements may be altered, on which channels the advert will be shown, and whether it may give the impression that the person is recommending a product or company.

This clarity is crucial, particularly where significant AI alterations are involved. Consent to the use of an original photograph does not necessarily constitute consent to a synthetically altered video in which the same person appears to make statements that they have never actually made.

The rights to the source material must also be clarified

The consent of the person depicted is only one aspect of clarifying rights. Furthermore, copyright may apply to the source material.

The Swiss Federal Institute of Intellectual Property (IGE) points out that copyright generally grants the rights holder the authority to decide whether, when and how a protected work may be used. Producers of sound and audiovisual recordings may also hold neighbouring rights.

Anyone who uses third-party photographic or video material and modifies it using AI must therefore check who holds the necessary rights of use and whether the existing licence also permits commercial use and editing. The fact that a video is freely accessible on a website, a platform or a social network does not mean that it can be freely used for one’s own advertising campaign.

The use of such material within an AI system may also raise copyright issues. The IGE points out that copyright issues arise both during the training of AI models and during their use. If protected works or parts thereof remain recognisable in the generated output, the output may, in particular, qualify as an adaptation. However, the legal development in this area is not yet complete.

It is therefore advisable for companies not to wait until the finished advert has been produced before carrying out a review. The clarification of rights should begin with the source material and explicitly take into account whether this may be processed using AI.

When AI feigns a recommendation

Advertisements in which AI creates the impression that a real person is endorsing a company, recommending a product or has made specific statements about a service are particularly sensitive.

In addition to the protection of personality rights, the law on fair trading may also be relevant. The Federal Act against Unfair Competition (UWG) covers, amongst other things, incorrect or misleading statements regarding business relationships. An advert that suggests a business relationship or recommendation that does not actually exist may therefore give rise to additional risks under competition law.

From a marketing perspective, the perceived authenticity of a well-known or professionally credible individual may be particularly appealing. Legally, however, it is precisely this impression that increases the risk if the person in question has never actually consented to the advertisement.

Satire is not a free pass for commercial AI content

Even a satirical or exaggerated portrayal does not automatically mean that personality rights are disregarded. Whilst it is recognised that satire and caricature may enjoy greater leeway due to their typical alienation and exaggeration, Whether a specific depiction is justified, however, remains a matter of weighing up the interests involved on a case-by-case basis.

Advertising campaigns should therefore not rely on the fact that labelling AI-generated content as ‘satire’ alone provides legal protection. In particular, where the depiction primarily serves to sell a product or promote a company, commercial interests must be weighed against the personal rights of the individual concerned.

What can a data subject do about unauthorised AI advertising?

If image or video material is used without sufficient consent, the data subject may take civil action against an unlawful infringement of their personal rights. Depending on the situation, options may include, in particular, preventing an imminent infringement, remedying an existing infringement and obtaining a court declaration of its unlawfulness. The FDPIC also expressly mentions the possibility of demanding the deletion or removal of images published without authorisation.

Financial claims are also possible, but are not automatically granted. Where there is demonstrable financial loss, compensation may be considered. In the event of a sufficiently serious infringement of personality rights, compensation for non-pecuniary damage may be due, subject to the statutory conditions.

If a company has made a profit from the unlawful use of a person’s personality rights, a claim for the surrender of profits may also arise under certain conditions. In particular, this requires an unlawful infringement of personality rights, a profit made, and a causal link between the infringement and that profit.

This last point can be of particular economic significance, especially in the case of successful online campaigns. However, the mere fact that an advertisement has generated turnover does not in itself mean that the entire economic gain must be surrendered. The link between the infringement of personality rights and the profit must be proven both in law and in fact.

Before the first AI prompt: clarify rights first

Companies, agencies and marketing departments should therefore treat AI-supported advertising in the same way as any other commercial media production. Before publication, it should be clarified who holds the rights to the source material, whether the person depicted has consented to the specific advertising use, and whether this consent also covers the planned alterations.

Equally important is the question of what impression the finished content creates on the audience. If new statements, actions or recommendations are attributed to a real person, the legal risk increases. A technically convincing AI edit can be even more problematic than an obviously artificial representation, precisely because the audience can barely distinguish between the original and the manipulation.

Contracts and consent forms should take this development into account. General clauses regarding the use of visual material are not always sufficient when it comes to far-reaching AI editing. The greater the scope for altering a person’s appearance, voice, statements or context, the more precisely it should be stipulated what is actually permitted.

Denmark as a potential pioneer in protection against deepfakes and Switzerland by way of comparison

Denmark intends to explicitly enshrine protection against deepfakes in law. A draft bill proposes amending the Danish Copyright Act to include a right of protection against realistic, digitally generated replicas of personal characteristics. The focus is on the face and voice. Those affected should, in particular, be able to take action against the unauthorised public dissemination of such content. This does not refer to a traditional copyright over one’s own face or voice. Rather, it is a right of protection rooted in personality rights, which is to be enshrined in the Copyright Act. Exceptions are envisaged, amongst other things, for caricature, satire, parody and pastiche – a distinction that is particularly crucial for journalism, advertising and the use of AI-generated content.

Switzerland does not currently have a specific legal offence relating to deepfakes. It relies on existing, technology-neutral provisions. The Federal Council recommended rejecting Motion 23.3563 Mahaim, which called for the regulation of deepfakes under civil, criminal and administrative law. Its reasoning: the current legislation contains no such loophole. Protection is provided in particular by the protection of personality rights under Article 28 et seq. of the Swiss Civil Code (ZGB), data protection law and, depending on the circumstances, criminal law provisions such as identity theft under Article 179decies of the Swiss Criminal Code (StGB). Anyone who clones a voice for a campaign or features a person in a synthetic video must therefore, under Swiss law, check whether this infringes the rights of the person concerned.

The preliminary draft of the Federal Act on Communication Platforms and Search Engines (KomPG) of 29 October 2025 will apply to the dissemination of such content. Modelled on the EU Digital Services Act, it provides for reporting procedures for unlawful content and further due diligence obligations for very large platforms. However, it does not create a separate legal right of protection against deepfakes.

The difference between the Danish and Swiss approaches therefore lies primarily in how deepfake protection is structured in law and how it can be enforced. In Switzerland, those affected must rely on various existing legal instruments. Limitations are particularly evident in the lack of post-mortem protection of personality rights and in the enforcement of claims against platforms. Denmark, by contrast, intends to create an explicit right of defence against realistic digital replicas. For businesses, the media and marketing departments, the specific use is key in both approaches: who is being replicated, has consent been given, and in what context is the content being published?

Conclusion: AI alters the material, not the rights

AI opens up new opportunities for businesses in terms of creative and personalised advertising. However, it does not eliminate personality rights, copyright or the requirements for fair advertising.

The most important practical rule is therefore: publicly accessible material is not automatically freely usable advertising material. Anyone who takes existing images or videos of real people, edits them using AI and then uses them commercially should clarify in advance both the rights of the person depicted and the rights to the source material.

This also remains relevant from a regulatory perspective. In Switzerland, there is currently no overarching legislation specifically covering artificial intelligence. Accordingly, the existing rules on personality rights, data protection, copyright and competition law apply in full to AI.

For businesses, it is therefore not decisive whether a campaign was produced with or without AI. What matters is whether they are using the content, individuals and rights in question lawfully.

Sources