EU AI Act ยท Article 50
EU AI Act Article 50: who must mark or disclose AI content
Last reviewed 17 September 2026. Regulation (EU) 2024/1689, Article 50. Last change: Article 50 applies, 2 August 2026.
Article 50 splits the work. Providers build transparency into the system; deployers tell the people exposed to it.
Providers
- 50(1)Interactive systemsPeople are told they are talking to AI, unless that is obvious
- 50(2)Generative systemsOutput marked machine-readable and detectable as AI-generated
Deployers
- 50(3)Emotion and biometric systemsInform the people exposed to them
- 50(4)Deep fakes and public-interest textDisclose that the content is AI-generated or manipulated
50(5): information must be clear and accessible by first exposure. 50(6) and (7): other law unaffected; codes of practice.
Article text, paragraph by paragraphRegulation (EU) 2024/1689, Art. 50
Providers must ensure that AI systems intended to interact directly with natural persons are designed and developed so that those persons are informed they are interacting with an AI system, unless this is obvious from the point of view of a reasonably well-informed, observant and circumspect person. Exceptions apply to systems authorised by law to detect, prevent, investigate or prosecute criminal offences, subject to safeguards.
Article 50(2)Providers of AI systems, including general-purpose AI systems, that generate synthetic audio, image, video or text content must ensure the outputs are marked in a machine-readable format and detectable as artificially generated or manipulated. The technical solutions must be effective, interoperable, robust and reliable as far as technically feasible, taking into account the specificities and limitations of content types, implementation costs and the generally acknowledged state of the art. The obligation does not apply where the system performs an assistive function for standard editing or does not substantially alter the input data or its semantics, or where authorised by law for criminal-offence purposes.
Article 50(3)Deployers of emotion recognition or biometric categorisation systems must inform the natural persons exposed to them, and process personal data in accordance with the applicable data-protection regulations, with a law-enforcement exception.
Article 50(4)Deployers of AI systems that generate or manipulate image, audio or video constituting a deep fake must disclose that the content has been artificially generated or manipulated, with exceptions for law-enforcement use and a modified duty for evidently artistic, creative, satirical or fictional work. Deployers that generate or manipulate text published to inform the public on matters of public interest must disclose this, unless the content has undergone human review or editorial control and a natural or legal person holds editorial responsibility.
Article 50(5)The information under paragraphs 1 to 4 must be provided to the natural persons concerned in a clear and distinguishable manner at the latest at the time of the first interaction or exposure, and must conform to accessibility requirements.
Article 50(6) and (7)Paragraphs 1 to 4 do not affect requirements in Chapter III (high-risk systems) or other Union or national transparency law. The AI Office encourages and facilitates codes of practice at Union level to facilitate the effective implementation of the detection and labelling obligations; the Commission may approve them or adopt common rules by implementing act.
Article 50(2) sets an obligation on providers of generating systems (marking, detectability). Article 50(1) and 50(4) set obligations on providers of interactive systems and on deployers (disclosure to the person). These are distinct duties held by different parties. The text does not name any particular technology and does not require identification of any natural or legal person in the mark.
Source: Regulation (EU) 2024/1689, Official Journal (EUR-Lex). These paragraphs summarise the article; the linked text is authoritative.
Common questions
Who must comply with Article 50 of the EU AI Act?
Providers and deployers, with different duties. Providers of AI systems that interact directly with people must ensure those people are told they are interacting with AI (50(1)), and providers of systems that generate synthetic audio, image, video or text must mark the output (50(2)). Deployers of emotion recognition or biometric categorisation systems must inform the people exposed to them (50(3)), and deployers of deep fakes or public-interest text must disclose that the content is AI-generated or manipulated (50(4)).
Does Article 50 require a visible AI label?
Not from the provider. Article 50(2) requires machine-readable marking. Disclosing to people that content is a deep fake is a deployer duty under 50(4), and under 50(5) the information must be clear and distinguishable at the latest at the first interaction or exposure.
Are there exceptions to the Article 50(2) marking duty?
Yes. It does not apply where the system performs an assistive function for standard editing, or does not substantially alter the input data or its semantics, or where use is authorised by law to detect, prevent, investigate or prosecute criminal offences.
Does Article 50 require a specific watermarking technology?
No. The text is technology-neutral: it names no technology and does not require the mark to identify any person. The technical solutions must be effective, interoperable, robust and reliable as far as technically feasible.
Other sections
How the Code of Practice marks content
The layered mark signatories commit to under 50(2), free detection, and the 2 February 2027 interoperability date.
The Code of Practice → Timeline and Article 99Dates and penalties
Article 50 applies from 2 August 2026; systems already on the market have until 2 December 2026. Fines up to EUR 15M or 3% of worldwide turnover.
Dates and penalties → Article 50(2)What 50(2) requires, and what it does not
Machine-readable, detectable marks from the provider. No specific technology, and no visible label from the provider.
What 50(2) requires → Factual profilesMarking and detection technologies
Tools profiled in the same fields, from each organisation's own materials. Not an endorsement.
Technologies →