Guidelines on the implementation of the transparency obligations for certain AI systems under Article 50 of the AI Act — C(2026) 5054 final, Annex

(126) The transparency obligation under Article 50(4), first subparagraph, AI Act will interact with the obligations of providers of VLOPs and VLOSEs to identify and mitigate systemic risks pursuant to Articles 34 and 35 DSA, in particular, Article 35(1)(k) DSA (35). The latter provision lists, among the possible risk-mitigation measures that providers of VLOPs and VLOSEs may adopt, measures ensuring that ‘an item of information, whether it constitutes a generated or manipulated image, audio or video that appreciably resembles existing persons, objects, places or other entities or events and falsely appears to a person to be authentic or truthful is distinguishable through prominent markings when presented on their online interfaces’. In addition, providers of VLOPs and VLOSEs may also provide an easy-to-use functionality which enables recipients of the service to indicate such information. While this DSA provision seems similar to Article 50(4) AI Act, it is complementary to it and contains two important distinctions:

- Different material scope of application: Compared to Article 50 (4) AI Act, Article 35(1)(k) DSA covers a broader range of content that may falsely appear authentic or truthful, regardless of the technologies used to create the content (i.e. it covers content that has been generated or manipulated by using AI systems or by any other tool).

- Different personal scope of application: Article 50(4) AI Act obliges deployers of AI systems generating or manipulating deep fakes to label such content, while Article 35(1)(k) DSA applies to providers of VLOPs and VLOSEs disseminating the content or making such content discoverable through search. For example, deployers of AI systems integrated into a VLOP under the AI Act may be considered ‘recipients of the service’ under Article 3(b) DSA (36) if they use the VLOP to disseminate content generated with the respective AI system, while the VLOP may be considered the provider of the AI system if the conditions of Article 3(3) AI Act are fulfilled. Where providers of VLOPs or VLOSEs make labelling tools available to such deployers enabling them to label their deep fake content in compliance with Article 50(4) AI Act (i.e. a clear and distinguishable disclosure of the AI-origin), those deployers can rely on such tools to fulfil their transparency obligation under that provision within the context of the VLOP or VLOSE used. Providing such a functionality is without prejudice to the responsibility of the deployers under the AI Act to fulfil their labelling obligations under Article 50(4) AI Act. Furthermore, providers of VLOPs and VLOSEs will also be considered deployers of an AI system under the AI Act if they use an AI-system under their authority for their own professional purposes (e.g. creating visuals to be used as part of marketing material).

(127) If the deep fake includes information relating to a directly or indirectly identifiable natural person (who is alive), then deployers as controllers under Union data protection law will also need to comply with the respective obligations under Union data protection

(35) See also Section 4.4 for further information regarding the interplay between Article 50 AI Act and the DSA.

(36) Recital 2 DSA.

law (37). This includes, for example, reliance on an appropriate legal basis for the processing of the personal data, information regarding the processing of data relating to the depicted data subjects, etc.

(128) Deployers of AI systems generating deep fakes will furthermore need to ensure that they comply with relevant Union intellectual property laws (such as Union trademark law and Union law on copyright or related rights) if their deep fake incorporates subject matter protected under those laws. The application of a label or icon to a deep fake does not, as such, have any impact on its eligibility for copyright protection, which should be assessed in accordance with applicable Union copyright law.

(129) Where existing persons are depicted in deep fakes, deployers will also need to comply with relevant legal regimes protecting personality rights over personal images or voices. The transparency obligation of Article 50(4), first subparagraph, AI Act does not imply that AI-generated or manipulated deep fakes that are harmful and unlawful under the applicable Union or national law (e.g. misleading advertising or criminal law such as child sexual abuse material and non-consensual intimate images) may be generated and disseminated (see also Section 4.4).