AI Act: The Commission provides guidance on the transparency obligations under Article 50
In May 2026, the European Commission published the highly anticipated draft guidelines (the "Draft Guidelines") on the implementation of the transparency obligations for certain AI systems under Article 50 of Regulation (EU) 2024/1689 (the "AI Act"). The consultation process has been completed, and the guidelines will likely be published before 2 August 2026, when Article 50 becomes applicable.
Article 50 of the AI Act applies to providers and deployers of certain AI systems and contains four separate transparency obligations for different types of systems.
Transparency for interactive AI systems
Based on Article 50(1) of the AI Act, providers must design AI systems that are intended to interact directly with natural persons in a way that ensures that the user is informed that they are engaging with an AI system.
The Draft Guidelines clarify the scope of Article 50(1), including that it only applies where the system involves a bidirectional exchange of information between the AI system and a natural person, which can occur in various forms (e.g. auditory, visual and physical) (para. 27). This will e.g. be relevant in relation to chatbots used for customer service or other AI systems where there is an exchange of information between the AI system and the user.
The transparency obligation does not apply if it is obvious that the user is engaging with an AI system. Based on the Draft Guidelines, which provide further details and examples of when this exception may be applicable (pages 13-15), the Commission appears to adopt a relatively strict interpretation of the exception.
Labelling of deep fakes
According to Article 50(4), first paragraph, of the AI Act, deployers of AI systems that generate or manipulate images, videos or audio constituting a deep fake must disclose that the deep fake content has been artificially generated or manipulated.
A "deep fake" is defined in Article 3(60) of the AI Act as "AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful".
Elaborating on this definition, the Draft Guidelines clarify that it is sufficient for simulated persons, objects, places, entities or events to resemble someone or something that can exist, or could have existed, in reality (realistic subjects), whereas content that defies the laws of nature or physics, or depicts lifeforms not commonly accepted in biology (such as dragons or elephants driving cars), falls outside the scope of the transparency obligation (para. 107).
As an example, if marketing material is generated using AI and depicts e.g. realistic landscapes, people etc., the material must include a disclosure. Contrary, AI-supported manipulation of minor, technical aspects of pre-existing content e.g. editing background details or audio noise reduction, does not render the resulting content to become a deep fake (para. 109).
The Draft Guidelines also provide guidance on the transparency obligation for evidently artistic, creative, satirical, fictional or analogous deep fakes under Article 50(4).
Additional transparency requirements under Article 50
In addition to the above, Article 50 of the AI Act and the Draft Guidelines also set out transparency obligations for providers of AI systems that generate synthetic audio, image, video or text content (Article 50(2)), requiring such outputs to be marked and detectable as AI-generated or manipulated, as well as transparency obligations for deployers of emotion recognition and biometric categorisation systems (Article 50(3)).
How to comply with the transparency obligations
Article 50(5) sets out requirements applicable to all transparency obligations under Article 50. Information must be provided in a clear and distinguishable manner, no later than the first interaction or exposure, and must comply with applicable accessibility requirements.
According to the Draft Guidelines, information is clear where it is noticeable and easy to understand, and distinguishable where it is easy to identify as separate from other information and the environment in which the content is presented. Information hidden in manuals or behind layers of menu options will generally not satisfy these requirements (para. 131).
As for disclosure under Article 50(1) (i.e. interactive AI systems such as chatbots), the transparency obligation could be fulfilled through written notifications (e.g. a chatbot that starts a conversation by mentioning that it is based on AI technology) (para. 33). However, techniques that would not fulfill the transparency obligation include disclosure contained only in terms and conditions (para. 35).
The code of practice on marking and labelling of AI-generated content
Based on Article 50(7), the European Commission's AI Office shall develop codes of practice to facilitate the effective implementation of the obligations regarding detection and labelling.
The Code of Practice on Transparency of AI-Generated Content (the "Code of Practice") was published on 10 June 2026.
While compliance with the Code of Practice is voluntary, it is intended to help providers and deployers fulfil their obligations under Article 50(2), (4) and (5) in terms of marking and labelling requirements and includes, among other things, uniform EU icons and text labels for deep fakes, guidance on placement etc.
Penalties
Based on Article 101 of the AI Act, providers and deployers that fail to comply with the applicable transparency obligations may be fined up to EUR 15,000,000 or, if the offender is an undertaking, up to 3% of its total worldwide annual turnover for the preceding financial year, whichever is higher.
Want to know more?
If you have any questions, please contact Plesner's Marketing and Consumer Law team.
Read the European Commission's Code of Practice on Transparency of AI-Generated Content