EU AI Act transparency guidelines issued

Frankfurt-based Dr. Nils Rauer, who specialises in AI-related law and regulation and co-leads Pinsent Masons’ global AI team, said some of the obligations have been open to interpretation and welcomed the European Commission’s attempts to clarify the requirements in its draft new guidance.

Under Article 50 of the AI Act, providers of AI systems face a general duty to ensure people they intend to interact directly with their system are informed that they are interacting with an AI system. That duty does not apply if it is “obvious from the point of view of a natural person who is reasonably well-informed, observant and circumspect, taking into account the circumstances and the context of use”. Exceptions also apply for certain law enforcement purposes.

Providers must also ensure that, where their system generate synthetic audio, image, video or text content, those outputs are marked in a machine-readable format and detectable as artificially generated or manipulated.

Deployers of an AI system that generates or manipulates image, audio or video content constituting a deep fake, must also generally disclose that the content has been artificially generated or manipulated, subject to limited law enforcement-related exceptions. If the content forms part of an evidently artistic, creative, satirical, fictional or analogous work or programme, the transparency obligations “are limited to disclosure of the existence of such generated or manipulated content in an appropriate manner that does not hamper the display or enjoyment of the work”.

Alexander Bibi, a member of Pinsent Masons’ global AI team, emphasised that where AI systems generate or manipulate text published with the purpose of informing the public on matters of public interest, deployers must further disclose that the text has been artificially generated or manipulated – again, subject to limited exceptions. Those exceptions include “where the AI-generated content has undergone a process of human review or editorial control and where a natural or legal person holds editorial responsibility for the publication of the content”, the AI Act states.

Businesses must also generally tell people when they are using an emotion recognition system or a biometric categorisation system.

The various transparency obligations must be discharged “in a clear and distinguishable manner at the latest at the time of the first interaction or exposure” and the information must “conform to the applicable accessibility requirements”.

Rauer said: “The Commission’s recently published guidance on Article 50 of the EU AI Act is a welcome attempt to bring clarity to some of the Act’s transparency concepts most open to interpretation. In particular, the clarification of when interactions with AI systems must be disclosed, and when they may be exempted because the AI nature is ‘obvious’, will be of immediate relevance for providers and deployers of chatbots and other interactive systems, especially in customer support contexts.”

“However, while the guidance offers a useful compass for lawyers, its practical value for non‑specialists remains limited, as for example assessments of ‘obviousness’ are anchored in familiar but abstract consumer‑law concepts such as the reasonably well‑informed, observant and circumspect ‘average consumer’. For smaller companies without in‑house legal expertise – those most reliant on off‑the‑shelf AI tools – translating these legal standards into concrete design and deployment decisions will remain challenging, raising the risk of inconsistent implementation and enforcement,” he said.

Bibi said tensions arise in the guidance around AI‑generated content and deepfakes.

“The Commission is candid in acknowledging that, at present, there is no single technical solution that can reliably mark and detect AI‑generated outputs across modalities, rendering the transparency obligation inherently theoretical for now,” Bibi said.

“While the more detailed explanation of what constitutes a ‘deepfake’ is helpful, it also confirms how wide the net is cast, as much AI‑generated imagery will fall within scope. This leaves deployers of AI‑generated advertising and other creative content facing finely balanced, case‑by‑case judgments as to whether a work is ‘evidently creative or fictional’, and if so, what form of disclosure is ‘appropriate’ without undermining user experience,” he said.

Bibi warned of the potential for litigation owing to uncertainties relating to disclosure, particularly, he said, since the guidance when finalised will be non-binding.

“Ultimately, it will fall to the EU’s highest court, the CJEU, to determine whether the Commission’s reading of Article 50 strikes the right balance between transparency, practicality and fundamental rights, and whether it can withstand the real‑world tension of rapid AI deployment,” Bibi said.

The draft guidelines issued by the Commission are open to consultation until 3 June 2026. The Commission has published a code of practice in parallel with the guidelines to support providers and deployers in meeting the specific transparency requirements around the marking and labelling of AI-generated outputs.