Transparency notices: Art. 50 disclosure duties
Track the Article 50 duties for systems that talk to people, read emotions, or generate synthetic content — chatbot disclosure, deepfake labelling, emotion-recognition notification — and generate the notices.
Documentation → Transparency Notices collects every Article 50 disclosure duty across the portfolio. These duties don't depend on risk tier — a minimal-risk chatbot still has to tell people it's a chatbot — so this workspace often has entries even when nothing you run is high-risk.
Which systems appear
The engine derives applicability from the behavioural flags you set when registering a system:
- Interacts with people (chatbot) → disclose that they're talking to an AI, unless it's obvious from context.
- Generates content → machine-readable marking of synthetic audio/image/video/text, and visible labelling of deepfakes.
- Emotion recognition or biometric categorisation → notify the people exposed to it.
- Placing content under your own name (AI-written text published to inform the public) → disclosure unless there's human editorial review.
If a system's flags change, regenerate its obligations and the transparency entries follow — see Classify a system.
Doing the work
Each row rolls up the system's transparency obligations. Click through to the system's Transparency tab to complete the checklist and store the wording you actually use: where the disclosure appears, the exact notice text, and evidence (a screenshot of the chatbot banner, the label spec for generated media).
The Art. 50 notice generator (Tools → generators, also linked from the tab) drafts the user-facing wording from your system's details — adjust the tone, then paste it into your product and attach the final version as evidence.
Deadline context
Art. 50 duties apply from 2 August 2026 — for most organisations the first hard AI Act deadline that touches ordinary products. The Calendar shows it against your registered systems; see Calendar & deadlines.