AI Act: Transparency must build trust – not undermine it

We strongly support the goal of clear and meaningful transparency for AI-generated deepfakes. But as currently drafted, the Guidelines risk going too far – with overly broad definitions and impractical labelling requirements that could lead to widespread over-labelling and damaging effects for professional media, esp. audio/radio.

Our key concerns

• The definition of “deepfake” should be narrowed to genuinely deceptive content instead of extending onto merely “realistic” content! Fictitious, non-deceptive content as well as minor content or technical adjustments should be explicitly excluded.
• Professional media already operate under strict editorial standards and regulatory oversight. Excessive labelling would not strengthen trust – it could erode it through visual clutter and “banner blindness.”
• Continuous or persistent labels risk confusing audiences rather than informing them. Besides that, labelling obligations must be proportionate and technically workable

Effective AI transparency should highlight real risks—not dilute them. VAUNET stands ready to continue the dialogue with the Commission to ensure that the Guidelines are clear, workable, and truly serve audiences.