EU AI Act Article 50 transparency obligations

Article 50 of Regulation (EU) 2024/1689 sets four transparency duties. Providers must tell people when they interact with an AI system (50(1)) and must mark synthetic audio, image, video and text output in a machine-readable, detectable way (50(2)). Deployers must notify people exposed to emotion recognition or biometric categorisation (50(3)) and must disclose deepfakes and AI-generated public-interest text (50(4)). Article 50(5) requires every disclosure to be clear, distinguishable and accessible at the first interaction or exposure.

Scan a product for Article 50 signals

Article 50 applies from 2 August 2026. The Regulation was published in the Official Journal on 12 July 2024 and entered into force on 1 August 2024. The European Commission's transparency guidelines and the Code of Practice on Transparency of AI-generated Content, both published in July 2026, explain how providers and deployers meet the obligations in practice. isGenAI records those dates on every scorecard and source link.

See the versioned source registry

Both. Article 50(1) and 50(2) bind providers, the organisations that build and place AI systems on the market. Article 50(3) and 50(4) bind deployers, the organisations that use those systems with people or publish their output. A company that builds and runs its own generative product carries both sets of duties, and Article 50(5) applies to every disclosure regardless of role.

Read the provider disclosure guide

Machine-readable marking means a signal that software can detect without a human reading the content: C2PA Content Credentials, provenance metadata such as IPTC digital source type, and robust watermarks are the mechanisms named in current practice. isGenAI's C2PA checker validates Content Credentials locally and reports signature validity and signer trust separately, and the provenance checker reads supported metadata, so you see exactly which marks a file carries and which transformations remove them.

Check a file for Content Credentials

Start with the free Article 50 scanner: enter a public product URL and get the observable disclosure signals, the controls each one maps to and the evidence each control still needs. Then upload generated files to the C2PA and provenance checkers to confirm machine-readable marking survives your export pipeline. For authenticated surfaces, submit authorized private workflow evidence through the scanner's intake flow. Every result records its method, date and limits so counsel can rely on it.

Open the free Article 50 scanner

No, and that is its strength. A scorecard is independent, externally observed evidence: public interaction disclosures, generated-artifact marking checks, provenance signals and change history, each with a status, confidence band and hash-linked evidence record. It is not legal advice and not an EU certification. It gives buyers, counsel and regulators verifiable facts, and ratings are never for sale.

Read the independence policy