Commission Guidelines on transparency obligations under Article 50 of the EU AI Act

Summary

The European Commission guidelines explain how to apply the transparency obligations in Article 50 of the AI Act. They clarify key concepts, exceptions, the marking and detection of AI-generated content, labelling of deep fakes and public-interest text, the clear and distinguishable information requirement, the role of a code of practice, penalties and the application date of August 2, 2026. The guidelines are non-binding and may be reviewed or amended.

Who it applies to

Non-binding guidance for competent authorities and for providers and deployers of AI systems that interact with people, generate or manipulate synthetic content, perform emotion recognition or biometric categorisation, or produce deep fakes and certain AI-generated text under Article 50 of the AI Act.

Key points

• Providers of interactive AI systems must ensure people are informed they are interacting with AI, unless the interaction is obvious or an exception applies.
• Providers of generative AI systems must mark outputs in a machine-readable format and make them detectable using effective, interoperable, robust and reliable solutions.
• Deployers must disclose deep fakes and certain AI-generated public-interest text, with a lighter regime for evidently artistic, creative, satirical or fictional works.
• Fines for infringements can reach €15 million or 3% of worldwide annual turnover, whichever is higher.
• Existing generative systems placed on the market before August 2, 2026 have until December 2, 2026 to meet the marking and detection obligations.

Key dates

Signed / adopted

Jul 20, 2026

Compliance date

Aug 2, 2026

Compliance date

Dec 2, 2026