AI transparency duties: what you have to disclose, and when
Article 50 is the rule most consumer AI touches. If your system talks to people, generates media or text, or reads emotions, you probably owe a disclosure. The exact duty depends on what the system does and whether you built it or deploy it. Pick your function below to see the subsection that applies, who carries it, and what a compliant disclosure looks like. These duties apply from 2 August 2026.
Interactive
Find your transparency duty
Choose the function that best describes your AI. This is guidance, not legal advice.
First, what is your role with this AI?
And what does your AI system do?
The four duties
The four transparency duties
Article 50 sets four disclosure obligations. Two fall on providers who build the system, two on deployers who use it.
Art. 50(1) · ProviderTell people they are dealing with AISystems built to interact with people must make clear a person is interacting with AI, unless that is obvious to a reasonably informed person.
Art. 50(2) · ProviderMark synthetic content as AI-generatedSystems that generate synthetic audio, image, video, or text must mark outputs in a machine-readable, detectable way.
Art. 50(3) · DeployerTell people about emotion or biometric systemsDeployers of emotion recognition or biometric categorisation must inform the people exposed to it.
Art. 50(4) · DeployerDisclose deepfakes and public-interest textDeployers must disclose deepfake content as artificially generated, and disclose AI-generated text published to inform the public, unless it went through human editorial review.
Questions
Common questions
Yes, in most cases. Article 50(1) requires that people interacting with an AI system are informed they are dealing with AI, unless it is obvious to a reasonably well-informed person. A clear line at the start of the chat is the standard way to meet it. There is a narrow carve-out for AI used by law enforcement to detect or investigate offences.
It splits. The provider that builds the system carries the Article 50(1) interaction disclosure and the Article 50(2) synthetic-content marking. The deployer that puts the system to use carries the Article 50(3) emotion and biometric notice and the Article 50(4) deepfake and public-interest disclosure. A single product can trigger duties on both sides.
2 August 2026, alongside the bulk of the Act. The AI Office is facilitating a code of practice under Article 56 to help with detection and labelling of synthetic content, so watch for that as the practical standard for how to mark outputs.
Get the disclosure wording and a labelling checklist
The checker names your duty. Members get ready-to-use disclosure language for chatbots, synthetic media, and deepfakes, plus a labelling checklist mapped to the Article 56 code. The checker stays free.
Grounded in Regulation (EU) 2024/1689, Article 50(1) to (4), Article 56, and Article 113. Educational resource, not legal advice. Last checked July 2026. Confirm the labelling standard against the Article 56 code before you build to it. New to the terms here? See the plain-English glossary. Full guide and sources.
EU AI Act Compliance Updates
Regulation changes, compliance deadlines, and enforcement actions — as they happen.
No spam. Unsubscribe anytime. We respect your data.