Transparency obligations (Article 50)

Set of information obligations of Article 50, allocated between provider and deployer according to the type of system: interaction with natural persons, synthetic content, emotion recognition or biometric categorisation, and deep fakes.

It is the only part of the Regulation that allocates obligations between whoever builds and whoever uses, paragraph by paragraph.

Which obligations it carries

All applicable from 2 August 2026, and the allocation by role is what to look at. Article 50(1) is an obligation of the provider: that systems intended to interact directly with natural persons are designed so that those persons know they are interacting with an AI system. Article 50(2) is an obligation of the provider: that the outputs of systems generating synthetic audio, image, video or text content are marked in a machine-readable format and detectable as artificially generated or manipulated; Article 111(4), added by Regulation (EU) 2026/1744, gives systems placed on the market before 2 August 2026 until 2 December 2026 to comply. Article 50(3) is an obligation of the deployer: informing the natural persons exposed to an emotion recognition or biometric categorisation system of its operation, and processing their data in accordance with the GDPR. Article 50(4) is an obligation of the deployer: disclosing that content constituting a deep fake has been artificially generated or manipulated. For the deployer, verification diligence over its provider’s Article 50(1) is an internal methodological control: there is no legal mandate requiring it.

What it is not

It is not a single obligation nor of a single party. Confusing the allocation is the most expensive error in this article: attributing 50(1) — the provider’s — to the deployer, or 50(4) — the deployer’s — to the provider, changes who is answerable. And it is not unconditional: Article 50(1) does not apply where this is obvious from the point of view of a natural person who is reasonably well-informed, observant and circumspect, taking into account the circumstances and the context of use; nor to systems authorised by law to detect, prevent, investigate or prosecute criminal offences, subject to the safeguards the paragraph itself requires.

The nuance almost nobody captures

The Article 50(1) exception — “where this is obvious” — is not a wide door: the standard is a reasonably well-informed, observant and circumspect person, judged in the specific context of use. A chatbot on a corporate website may be obvious to someone looking for it and not to someone arriving from a link. The exception is assessed by context, not declared once.

To find out more

• The Digital Omnibus is now in force: what changes in the AI Act and what stays the same
• The 'you're talking to an AI' notice is the provider's job, not yours
• The article that didn't move: why 2 August is still a real date
• Who must warn that your chatbot is an AI? The nuance in Article 50(1) that almost everyone gets wrong
• A well-designed chatbot: disclosure, a human way out, and the choice not to use it
• Do you have to label all AI-written text? The editorial review route
• Article 50 transparency, in application since 2 August 2026
• Before 2 August: what to close out on AI transparency