EU finalizes AI transparency guidance

The European Commission has published final Guidelines on transparency obligations under the EU AI Act, which largely apply from 2 August 2026. The guidance expands on the draft issued in May and clarifies territorial reach, cumulative obligations across a single AI system, and when providers, deployers, intermediaries and open-source providers are covered.

Providers outside the EU may fall within scope when their AI system outputs are used in the EU, while deployers outside the EU may be covered where they foresee dissemination and use in the EU. Interactive AI systems must inform people that they are dealing with AI, and the exception for obvious interactions is framed narrowly. AI agents that may interact with people must disclose both their artificial nature and the person on whose behalf they act.

The guidance adds detail on machine-readable marking and detection of AI-generated or manipulated content, including where AI agents produce perceptible synthetic outputs. It also takes a strict view of deepfakes, covering realistic replicas, avatars and AI-generated personas where content could falsely appear authentic or truthful. The detection and marking obligation for AI systems put on the market before 2 August 2026 applies from 2 December 2026, and breaches may carry fines of up to €15 million or 3% of annual worldwide turnover, whichever is higher.