Marketing teams routinely use artificial intelligence tools to create images, videos and voice tracks. Since 2 August 2026, Article 50 of Regulation (EU) 2024/1689 on artificial intelligence, the AI Act, has required disclosure that certain outputs have been generated or manipulated using AI. The July 2026 amendment, Regulation (EU) 2026/1744, did not postpone that date for businesses using AI. Only providers of generative systems placed on the market before 2 August 2026 received a transitional period until 2 December 2026, covering machine-readable marking of outputs under Article 50(2), as provided by amended Article 111(4).
A business using AI in marketing is usually a deployer
The AI Act distinguishes a provider, which develops or places an AI system on the market, from a deployer. A deployer is anyone using an AI system in the course of their activities other than purely personal use (Article 3(4)). An online shop or agency generating visuals with an existing tool is a deployer. Our article on roles under the AI Act explains when a user becomes a provider. Informing customers about a chatbot is a separate obligation, addressed in our guide to website chatbots and the AI Act.
When content is a deepfake
The Regulation defines a deepfake as follows.
Original statutory wording:
„deep fake“ je obrazový obsah, audioobsah alebo videoobsah vytvorený alebo zmanipulovaný pomocou AI, ktorý sa podobá na existujúce osoby, objekty, miesta, subjekty alebo udalosti a osobe by sa falošne javil ako autentický alebo pravdivý
Unofficial translation of the quoted Slovak wording: A deepfake is AI-generated or manipulated image, audio or video content that resembles existing people, objects, places, entities or events and would falsely appear authentic or truthful to a person.
In marketing, examples include a video in which a real public figure appears to speak, a cloned voice in a radio advertisement or a realistic photograph of a product in a place where it has never been. What matters is whether a person could perceive the content as real. In our view, an obviously drawn or stylised illustration will generally not be a deepfake. Article 50 contains a separate rule for text, which is outside this definition.
A deployer creating or manipulating such content using AI must disclose that it has been artificially generated or manipulated (Article 50(4)). The information must be clear and distinguishable and appear no later than the person’s first exposure to the content (Article 50(5)). In its questions and answers on Article 50, last updated on 24 July 2026, the European Commission states that the disclosure should be understandable and perceptible without technical aids, for example through a visible or audible notice. According to the Commission, a deployer cannot rely solely on machine-readable marking embedded by the tool provider under Article 50(2).
Where the content forms part of an evidently artistic, creative, satirical, fictional or similar work, the obligation is limited to appropriate disclosure of its existence in a way that does not hamper the display or enjoyment of the work. It does not disappear altogether. Whether an advertisement with a humorous ending qualifies as an evidently satirical work will depend on the particular circumstances.
AI-generated text
The obligation is narrower for text. Its origin must be disclosed when it is published to inform the public about matters of public interest. The obligation does not apply if the text has undergone human review or editorial control and a natural or legal person bears editorial responsibility for its publication (Article 50(4)). In our view, an ordinary product description or advertisement will generally not concern a matter of public interest. Articles about health or product safety on a company blog may be different. You can use the human review exception if someone with the necessary knowledge checks and approves the substance of the text before publication and the company takes responsibility for publishing it. Correcting spelling or giving formal approval alone is insufficient, as the European Commission explains in its answers on Article 50.
Personal consent and advertising disclosure
A deepfake of a real person must also respect personality rights. Likenesses, images and visual or audio recordings relating to an individual may be made or used only with that person’s consent (§ 12(1) of the Slovak Civil Code). The exceptions in § 12(2) and (3) concern official, scientific and artistic purposes and news reporting. The Act does not list advertising among them. An “AI-generated” label does not replace missing consent. When working with content creators, consent to use a face and voice, including AI alterations, should be agreed in the contract, as discussed in our article on licensing influencer content.
AI disclosure is separate from advertising disclosure. A paid post must be recognisable as advertising regardless of whether AI was involved in creating it (advertising disclosure for influencers).
For breaches of Article 50, the Regulation sets a maximum administrative fine of EUR 15,000,000 or 3% of total worldwide annual turnover, whichever is higher (Article 99(4)). Small and medium-sized enterprises are subject to the lower of those figures (Article 99(6)). For breaches of transparency obligations, the lower figure also applies to businesses in the small mid-cap category, known as SMCs (Article 99(6a)). Member States determine the specific rules for imposing penalties (Article 99(1)).
What to put in place in marketing
We recommend introducing the following rules within the marketing team:
- List the AI tools used by the team and external agencies and the content they are used for.
- Decide how you will label deepfakes. For images and video, visible text within the image is one option. For audio, use a notice at the beginning of the recording.
- Introduce substantive review of AI-generated text before publication and record who checked and approved it.
- Obtain consent before using a real person’s likeness or voice. Keep it in writing for evidential purposes.
- Specify in the agency or content creator contract who labels the content and who is responsible if the label is missing.
How we can help
We can establish an inventory of tools, labelling rules and internal responsibilities through our AI Act compliance service. We can include consent to use likenesses and voices and the allocation of responsibilities in an influencer agreement. Our copyright and licensing service addresses rights in AI-generated content.
This article provides general legal information as at 16 October 2026. It does not constitute legal services or advice on your specific matter. Laws change and the details of your situation may differ. Check the appropriate course of action or contact us before making a decision.
