Regulation (EU) 2024/1689

Article 50, explained clearly.

A practical introduction to the EU AI Act transparency obligations for specific providers, deployers, and use cases—not a substitute for legal advice.

Applicable from 2 August 2026

Interaction disclosure — Article 50(1)

Providers of systems intended to interact directly with people must ensure they are informed that they are interacting with AI, unless this is obvious in the circumstances and context. The information must be clear, distinguishable, accessible, and available by the first interaction.

Practical implementation

Place a plain-language notice in or immediately before the AI interaction. A generic “AI-powered” marketing claim is not automatic proof that the requirement is met.

Start with scope

Map the system, its intended use, the provider and deployer, audience, and possible exceptions. A website alone rarely proves all of those facts.

Synthetic content marking — Article 50(2)

In-scope providers generating synthetic audio, image, video, or text must make outputs machine-readable and detectable as artificially generated or manipulated. Measures should be effective, interoperable, robust, and reliable as far as technically feasible.

What it does not mean

C2PA is a useful provenance approach, but it is not mandated by Article 50. A visible label or HTML metadata alone does not prove output-level marking.

Questions to document

Which outputs are in scope? How are they marked? How is detectability tested? Does an assistive-editing or other exception apply?

Biometric systems — Article 50(3)

Deployers of emotion-recognition or biometric-categorisation systems must inform people exposed to their operation and comply with applicable data-protection rules.

Deepfakes and public-interest text — Article 50(4)

Deployers must disclose deepfake image, audio, or video. AI-generated or manipulated text published to inform the public on matters of public interest must also be disclosed unless human review/editorial control and responsible editorial ownership apply. Context and exceptions matter.

What SentinelGo can and cannot establish

SentinelGo records public-page evidence and potential gaps. It cannot conclusively establish the responsible actor, intended use, authenticated user journey, output-level marking, editorial control, timing of first exposure, or a legal exception. Use the report as a review checklist, not a legal determination.

Collect evidence before you decide.

Run a public-page screening to find observable transparency evidence and the questions your team should answer.

Start free audit