To help providers and deployers meet the transparency requirements for AI systems under Article 50 of the AI Act, the Commission has published draft guidelines for feedback. The draft guidelines take into account the input from previous consultations.
How to participate
Stakeholders can take part in this targeted consultation until 3 June 2026. For transparency, you can download the consultation questions below.
To ensure a fair and transparent process, only responses received through the online questionnaire will be considered and reflected in the final summary report.
This survey targets companies, ranging from startups and SMEs to large companies, and other organisations that develop and deploy AI systems that interact with individuals or generate synthetic content, including deep fakes. Stakeholders, including providers and developers of AI systems, businesses and public authorities as well as academia, research institutions and citizens are invited to share their views.
Background and objectives
The guidelines on transparency will clarify the scope and help deployers and providers of interactive and generative AI systems to comply with their respective transparency obligations.
The rules will become applicable on 2 August 2026. Providers of AI systems will have to inform users when they are interacting with an AI system and implement machine-readable marks in generative AI systems to enable the detection of synthetic content as AI generated or manipulated. The rules will also require deployers to inform people when they are exposed to deep fakes and AI-generated publications on matters of public interests and emotion recognition or biometric categorisation systems.
These guidelines complement the Code of Practice on transparency of AI-generated content which is also being finalised in parallel. The Code will be a voluntary tool to support providers and deployers of generative AI systems in the practical and effective implementation of the AI Act’s marking and labelling obligations for AI-generated and manipulated content.