Article 50 • transparency
Article 50 is now an active EU AI Act compliance requirement.
From 2 August 2026, specified providers and deployers must implement transparency measures for direct AI interaction, synthetic content, deepfakes, emotion recognition, biometric categorisation and certain AI-generated public-interest text.
Updated and legally reviewed: 26 September 2026.
Who must do what
Article 50 contains different obligations for different actors and use cases.
Direct interaction with AI
Providers of AI systems intended to interact directly with natural persons must generally design the system so people are informed they are interacting with AI unless that fact is obvious from the circumstances.
Synthetic content marking
Providers of systems generating synthetic audio, image, video or text content face machine-readable marking and detection-related requirements under Article 50(2), subject to the legal details and applicable technical state of the art.
Deepfakes
Deployers generating or manipulating image, audio or video that constitutes a deepfake must disclose that the content has been artificially generated or manipulated, subject to the Article's qualifications.
Public-interest text
Deployers of AI-generated or manipulated text published to inform the public on matters of public interest have disclosure duties unless the content has undergone qualifying human review/editorial control as provided by the law.
Transition
The general start date is 2 August 2026 — with one limited grace period.
For AI systems placed on the market before 2 August 2026, the Article 50(2) marking/detection obligation has a transition until 2 December 2026. The Commission states that content created before 2 August 2026 does not need retroactive labelling.
Implementation controls
What an Article 50 readiness review should test
System classification
Which systems interact with people, generate content, perform emotion recognition/biometric categorisation, or create material that can fall within deepfake/public-interest rules?
Provider vs deployer duties
Do not collapse the obligations. The party providing the system and the party deploying generated content can face different requirements.
User disclosure
Is the disclosure timely, clear and tied to the interaction rather than buried in general terms?
Technical marking
Can relevant generated/manipulated output be marked in a machine-readable way and detected as required, with technical limitations documented?
Content workflow
Do marketing, communications, media and product teams know when deepfake/public-interest labelling rules are triggered?
Evidence
Maintain system inventory, disclosure design, technical implementation, testing, exceptions/qualifications and change records.
Transparency Code
The Code of Practice is voluntary, but operationally important.
The Commission and AI Board assessed the Code of Practice on Transparency of AI-generated Content as an adequate voluntary instrument for Article 50(2), (4) and (5). The Commission reported about 190 organisations signed by the end of July 2026. Adherence is not conclusive proof of compliance, but it can provide a structured implementation pathway.
Official guidance