AER has joined a coalition of European audio and audiovisual media organisations – including VAUNET and VOEP – in calling for targeted improvements to the European Commission’s draft Guidelines on the transparency obligations under Article 50 of the AI Act.
The joint letter, addressed to Executive Vice-President Henna Virkkunen, welcomes the Commission’s efforts to provide greater legal clarity on the new AI transparency rules. At the same time, it raises serious concerns that parts of the current draft Guidelines risk producing outcomes that are disproportionate, technically unworkable and ultimately counterproductive for professional media.
The concern is not with the principle of transparency itself. Media organisations fully support the need for clear information where AI-generated or manipulated content creates a genuine risk of misleading audiences. The problem lies in the way the draft Guidelines currently interpret the relevant provisions of the AI Act. As drafted, they risk extending labelling obligations far beyond deceptive or high-risk content and into a wide range of professional editorial material that is clearly fictitious, harmless or only minimally adjusted. According to the signatories, this would lead to systemic over-labelling, legal uncertainty and significant operational burdens for media providers.
Together with the co-signatories, AER is calling for three targeted adjustments to the draft Guidelines:
- A narrower and clearer definition of “deepfake”
The current draft risks capturing content that is merely “realistic”, even where there is no genuine risk of deception. The joint letter calls on the Commission to limit the notion of deepfake to content that presents a real risk of misleading audiences, and to explicitly exclude fictitious, non-deceptive content as well as minor technical or editorial adjustments.
- Clarification of the timing of disclosure obligations
The signatories argue that Article 50(5) of the AI Act sets the latest point at which disclosure must be made, but does not create an obligation for continuous or persistent labelling throughout a piece of content. The current wording of the draft Guidelines risks implying precisely that, which would be disproportionate and, in many cases, impractical.
- Proportionate and workable labelling standards
The letter also calls for labelling requirements that reflect technical realities and human perceptual limits. Excessive or repeated labelling across large volumes of low-risk content would not strengthen trust in transparency; on the contrary, it could lead to “banner blindness” and weaken the impact of warnings in the cases where they are genuinely needed.
For radio and audio services, the issue is especially sensitive. Unlike visual media, audio formats do not lend themselves easily to repeated or persistent disclosures without affecting the listening experience itself. Requiring frequent or continuous transparency notices in low-risk cases would risk interrupting programming, confusing listeners and ultimately reducing the effectiveness of the disclosure.
Transparency measures should help audiences identify genuine risks of deception – not create warning fatigue or desensitise listeners through repeated notices attached to content that is not misleading in the first place. The joint letter calls on the Commission to adjust the draft Guidelines so that they remain faithful to the AI Act’s purpose: building trust through meaningful transparency, rather than undermining it through excessive and indiscriminate labelling.
Read the joint letter here.
For more information, please contact the AER office in Brussels at aer(Replace this parenthesis with the @ sign)aereurope.org.