Telling people they are talking to a machine
Article 50 adds transparency duties regardless of risk tier: systems interacting with people must disclose it unless obvious; synthetic audio, image, video and text must be marked in a machine-readable format; deep fakes and AI-generated text on matters of public interest must be disclosed; and emotion recognition or biometric categorisation must be declared to the people exposed to it.
Why it is like this
These are the harms that come from not knowing, rather than from the decision itself. Disclosure is a cheap fix, so it applies broadly rather than only to high risk.
Not to be confused with
A watermark requirement only. The marking duty is on the provider and must be machine-readable and interoperable; the disclosure duty for deep fakes is on the deployer, and is about what the audience sees.
In practice
A marketing team publishing an AI-generated article on an election has a disclosure duty even though the system is minimal risk.
Regulation (EU) 2026/1744 gave systems already on the market before 2 August 2026 until 2 December 2026 to meet the machine-readable marking duty. Everything else in Article 50 applied from 2 August 2026.
Source: Regulation (EU) 2024/1689 (AI Act) — Article 50. Read on .
Predict Check yourself
A minimal-risk chatbot generates synthetic images. Which Article 50 duties apply?
Show the answer
Both — disclose that the user is interacting with an AI system, and mark the output as artificially generated in a machine-readable format.
Article 50 cuts across the tiers. The "obvious" exemption covers the interaction disclosure, never the marking of synthetic content.