11 · Transparency Obligations (Article 50)
- Regulation (EU) 2024/1689, Article 50 + Commission Article 50 guidelines + Code of Practice on Transparency of AI-Generated Content + Regulation (EU) 2026/1744
- Applicable since: 2 August 2026, with a limited Article 50(2) transition until 2 December 2026 for systems placed on the market before 2 August 2026
- Last verified: 2026-09-07
Article 50 assigns different transparency duties to providers and deployers. It does not create one general duty to label every AI-assisted output.
Provider obligations
Direct interaction with an AI system (Article 50(1))
Providers must design AI systems that interact directly with natural persons so people are informed that they are interacting with AI, unless this is obvious to a reasonably well-informed, observant and circumspect person.
The Commission guidelines say the notice should be clear, distinguishable, accessible and given at the start of the first interaction. Systems operating only in the background, through machine-to-machine communication or through a human intermediary fall outside this specific duty.
Machine-readable marking of synthetic content (Article 50(2))
Providers of AI systems, including GPAI systems, that generate synthetic audio, image, video or text must ensure that outputs are marked in a machine-readable format and detectable as AI-generated or manipulated. The technical solution must be effective, interoperable, robust and reliable as far as technically feasible.
The guidelines identify exclusions and limits, including:
- short sequences of numbers, symbols or letters;
- source code;
- machine-to-machine outputs not exposed to people;
- non-final outputs used only in closed-loop industrial or production environments;
- assistive functions for standard editing; and
- a narrow business-to-business or industrial-context exemption where the guideline conditions are met.
Deployer obligations
Emotion recognition and biometric categorisation (Article 50(3))
Deployers must inform natural persons exposed to an emotion recognition or biometric categorisation system about the system's operation. The duty applies to real-time and retrospective use.
Deepfakes and public-interest text (Article 50(4))
Deployers must clearly disclose deepfake image, audio or video content no later than first exposure. A machine-readable provider mark does not by itself satisfy this human-facing disclosure duty.
Deployers must also label AI-generated or manipulated text published to inform the public on matters of public interest. The labelling duty does not apply where the text has undergone substantive human review or editorial control and a person or entity holds editorial responsibility. Spell-checking or grammar correction alone is not substantive review.
Artistic, creative, satirical and fictional works receive a limited disclosure treatment that must not hamper display or enjoyment of the work.
Form of the disclosure (Article 50(5))
Information required by Article 50 must be clear and distinguishable and must meet applicable accessibility requirements. The disclosure must reach the person exposed to the AI system or content, not remain available only through a technical metadata inspection.
Transparency Code of Practice
The Commission published the final Code of Practice on Transparency of AI-Generated Content on 10 June 2026. It has two sections:
- provider measures for machine-readable marking and detection under Article 50(2); and
- deployer measures for labelling deepfakes and public-interest text under Article 50(4).
The Commission and AI Board assessed the Code as an adequate voluntary tool on 9 July 2026. Signatories can rely on its measures to demonstrate compliance with the marking and labelling duties. Organisations that do not sign must demonstrate compliance through other adequate means.
By the end of July 2026, about 190 organisations had signed the Code. The Commission maintains the official signatory list; organisations can still sign after the initial list.
The Code does not replace Article 50 or the Commission guidelines and does not create a presumption of conformity. The EU also published optional icons that deployers may use for human-facing labels.
Timing and transition
| Date | Effect |
|---|---|
| 2 August 2026 | Article 50 applies |
| 2 December 2026 | Providers of synthetic-content systems placed on the market before 2 August 2026 must comply with Article 50(2), under Regulation (EU) 2026/1744 |
Content generated before 2 August 2026 does not require retroactive labelling, although the Commission encourages voluntary labelling where practical.
Enforcement
National market surveillance authorities are the main enforcers. The AI Office has a narrower role for systems within Article 75 supervision, and the European Data Protection Supervisor enforces the rules for EU institutions, bodies and agencies.
Key references
| Topic | Reference |
|---|---|
| Direct AI interaction notice | Art. 50(1) |
| Machine-readable marking and detection | Art. 50(2) |
| Emotion recognition and biometric categorisation notice | Art. 50(3) |
| Deepfake and public-interest text labelling | Art. 50(4) |
| Clear, accessible disclosure | Art. 50(5) |
| Code of Practice | Art. 50(7) |
| Pre-August 2026 system transition | Art. 111(4), as inserted by Regulation (EU) 2026/1744 |