Transparency (Art. 50)
Article 50 of the EU AI Act imposes transparency duties that cut across risk tiers: even a minimal-risk chatbot owes a disclosure. AI SENTINEL tracks the four Art. 50 obligations per AI system, splits them by who actually owes them, and counts down the one deadline that has a transitional period.
When it applies
Art. 50 has applied since 2 August 2026 (Art. 113). Regulation (EU) 2026/1744 — the Digital Omnibus on AI, OJ L of 24 July 2026 — grants systems placed on the market before that date a transitional period for the machine-readable marking duty in Art. 50(2), running to 2 December 2026.
That grace period is the reason the AI Registry asks whether a system was placed on the market before 2 August 2026. Answer it accurately: it is the only input that moves the marking deadline, and the dashboard's “marking overdue” counter is derived from it.
The four obligations
Two are owed by the provider and two by the deployer. A single organization often holds both roles, but recording them separately is what lets you answer a regulator asking which hat you were wearing.
AI interaction disclosure
Systems that interact directly with people must make clear they are AI, unless that is obvious to a reasonably well-informed person. Chatbots and voice agents are the typical case.
Machine-readable marking of synthetic content
Systems generating synthetic audio, image, video, or text must mark their output in a machine-readable format detectable as artificially generated. This is the obligation with a transitional deadline — see below.
Emotion recognition and biometric categorisation
Deployers must inform the people exposed to an emotion-recognition or biometric-categorisation system that it is in use. Note that several such uses are prohibited outright under Art. 5 — the module flags that overlap rather than letting you record a disclosure for a practice you cannot lawfully run.
Deepfake and public-interest text labelling
Deployers publishing deepfakes must disclose that the content is artificially generated or manipulated. The same applies to AI-generated text published to inform the public on matters of public interest.
Recording status
Not applicable
This obligation does not apply to the system. The default for every obligation until you decide otherwise.
Required
The obligation applies and is not yet satisfied. Systems in this state drive the deadline counters on the executive dashboard.
Implemented
The obligation applies and the disclosure or marking is in place. Record how in the AI statement so the evidence survives staff turnover.
Working through it
- 1
Open the system's Transparency tab
Every AI system in the registry has a Transparency (Art. 50) tab alongside Models, Data Sources, and Risk Classification.
- 2
Review the suggested obligations
The module screens the system's purpose and technique and suggests which of the four obligations look applicable. Treat these as prompts, not conclusions — you confirm each one.
- 3
Set a status per obligation
Mark each as Not applicable, Required, or Implemented. Anything left Required appears in the dashboard's transparency counters.
- 4
Write the AI statement
Record the disclosure wording and the marking method you actually use. This is the artifact a market surveillance authority will ask for; a status field alone will not satisfy them.
A note on scope
This module records your Art. 50 position; it does not mark content for you. Machine-readable marking is implemented in the generating system itself. AI SENTINEL is where you document which systems owe the duty, whether it is satisfied, and how — so the answer exists before someone asks for it.