CANCELLED - Beyond the binary label: implementing meaningful AI disclosures under Article 50 of the EU AI Act

• Robert Schuman Centre for Advanced Studies
• EGPP: European Governance and Politics Programme

Article 50(4) of the EU AI Act mandates clear rules for AI transparency, requiring providers and deployers of AI to notify citizens when they are interacting with deepfakes and AI-generated or modified text on matters of public interest.

This presentation builds on the implementation of this transparency principle through the Code of Practice on AI (drafted as a roadmap for compliance with the obligations under the AI Act), and reflects on how the technical presence of discourses can remain meaningful for users.

Drawing on empirical testing of three visual icon variants in France and Romania, the analysis underlines a critical dichotomy: while AI-related labels enhance user awareness, boosting recognition by 22 % and achieving almost 70% credibility, almost half of the participants still failed to identify the content they were exposed to as synthetic.

These findings show that binary signals (e.g., AI-generated) are insufficient to address the deceptive realism of synthetic content. To address this, the Task Force of the Code on AI - in which the author of this presentation is involved - proposes a progressive, two-layered transparency focused on contextualised disclosure. Finally, the analysis discusses behavioural implications for citizens and suggests a framework for translating compliance with regulation into accessible, citizen-centric transparency.

At the EUI and the Robert Schuman Centre, we are dedicated to removing barriers and providing equal opportunities for everyone. Please indicate in the registration form your accessibility needs, if any. Alternatively, you can contact the logistics organiser of the event.