EU AI Act transparency duties take effect with limited deferral

Article 50 of the EU AI Act now requires covered organizations to disclose certain uses of AI, including when individuals interact directly with AI systems, when emotion-recognition or biometric-categorisation systems are used, and when specified content has been generated or manipulated by AI.

The obligations differ by role. Providers that place or put AI systems into service in the European Economic Area must notify individuals when direct interaction with AI is not obvious and ensure synthetic text, images, audio, or video are marked in a machine-readable format where required. Deployers using AI systems must inform individuals exposed to emotion-recognition or biometric-categorisation tools and provide disclosures for deepfakes and certain public-interest text.

The EU Digital Omnibus deferred only a limited provider obligation. The four-month transition period applies to machine-readable marking and detection for in-scope systems, including general-purpose AI systems, that were placed on the EEA market before 2 August 2026, giving providers until 2 December 2026 to comply. Deployer duties have applied since 2 August 2026.

Businesses are advised to determine whether they are providers or deployers, assess whether Article 50 applies to each AI use case, document any reliance on the transition period, and build transparency into notices, procurement, contracting, and content approval processes.