In short. AI Act Article 50 requires transparency about certain AI interactions and outputs, with separate duties for providers and professional users.
Definition: Article 50 of the EU AI Act sets transparency obligations for providers and professional users of certain AI systems. These include informing people about direct AI interaction, marking synthetic outputs in a machine-readable format, and disclosing deepfakes and certain AI-generated text publications.
Key facts
- As of 17 October 2026, the Commission's Article 50 FAQ gives an application date of 2 August 2026. For systems already on the market before that date, the provider's marking and detection obligation has a transition until 2 December 2026.
- The Commission's legal-text mirror assigns machine-readable marking to providers and public-interest text disclosure to deployers, meaning people or organisations using AI systems under their authority professionally.
- The Commission's transparency guidelines require substantive human review or editorial control, alongside editorial responsibility, for the text disclosure exception. Spell-checking alone falls short.
- The Commission describes its Code of Practice on Transparency of AI-generated Content as a voluntary way to demonstrate compliance with the relevant legal obligations.
Who carries the obligations?
A provider develops an AI system, or has it developed, and supplies it under its own name or trademark. A deployer uses a system under its authority for professional activities. The Commission's FAQ explains that an advertising company can remain the deployer when contractors operate the system under its responsibility and control.
The legal text separates the duties by activity. Providers of systems interacting directly with people must inform them of that interaction, unless it is obvious in context. Providers of systems generating synthetic audio, images, video or text must mark outputs in a machine-readable format and enable detection, subject to specified exceptions, including standard editing assistance. Deployers also have information duties for emotion recognition and biometric categorisation systems.
The relevant legal identifiers are:
Regulation (EU) 2024/1689
Article 50(4), second subparagraph: public-interest text disclosure
Article 50(5): timing and accessibility of information
Which website publications require disclosure?
Article 50(4) covers AI-generated or manipulated text published to inform the public on matters of public interest. The Commission's guidelines give a newspaper's AI-generated summary about a town council decision as an example. They also include articles discussing diets and disease, and corporate reports containing investor information.
The same guidelines place company advertisements and product descriptions outside this text category when they contain no claims concerning, for example, health, consumer safety or sustainability. The purpose and substance of a publication therefore matter when assessing its disclosure duties.
The text exception has two cumulative conditions: human review or editorial control, and a person or organisation holding editorial responsibility for publication. The guidelines describe review as deliberate examination by people with relevant knowledge and professional judgement. Fact-checking is a minimum requirement. Editorial responsibility means ultimate legal responsibility for publication, including the review or control process.
How must the disclosure appear?
Under Article 50(5), information must be clear and distinguishable by the first interaction or exposure, and meet applicable accessibility requirements.
A deepfake is AI-generated or manipulated image, audio or video that resembles existing people, objects, places, entities or events and falsely appears authentic or truthful. The Commission's FAQ requires disclosure perceivable without special tools or dedicated actions, such as a visible or audible label. The provider's embedded machine-readable mark alone is insufficient for the deployer's deepfake disclosure duty. Evidently artistic or fictional works have an adapted disclosure requirement that preserves their display or enjoyment.
Related terms
FAQ
Does checking spelling qualify for the text exception?
The Commission's FAQ excludes superficial spelling and grammar checks. You need substantive human review or editorial control and an identified holder of editorial responsibility.
What if AI rewrites an article after editorial approval?
The guidelines state that substantive AI intervention after review makes the exception void. If you rely on the exception, the resulting text needs the required review or editorial control before publication.
Is signing the transparency code compulsory?
The Commission's code page makes participation voluntary. If you comply through other means, you must demonstrate that those measures are adequate.
Sources
- European Commission, "Article 50: Transparency obligations for providers and deployers of certain AI systems", consolidated text, 27 July 2026
- European Commission, "Transparency obligations under Article 50 of the AI Act", 24 July 2026
- European Commission, "Guidelines on the implementation of the transparency obligations for certain AI systems under Article 50 of Regulation (EU) 2024/1689", 20 July 2026
- European Commission, "Code of Practice on Transparency of AI-generated Content", accessed 17 October 2026
Check how a page reads to answer engines. Open the GEO score analyzer