EU begins enforcing AI Act transparency rules

The European Commission and EU national authorities have begun enforcing key transparency obligations under the EU AI Act, with new rules entering into application on 2 August 2026. The measures aim to improve accountability by ensuring people are informed when they interact with AI systems or encounter AI-generated content.

Under the new requirements, providers of chatbots and other AI systems that communicate directly with users must generally disclose that users are interacting with artificial intelligence, unless this is already clear. Systems that generate or modify text, audio, images or video must also make their outputs detectable as AI-generated through machine-readable information. Additional disclosure duties apply to organisations creating or distributing deepfakes and AI-generated content related to matters of public interest.

The rules do not require identical labels for all AI-assisted content. Obligations depend on the role of an organisation, whether it develops an AI system, deploys it, or publishes AI-generated outputs. More than 180 organisations have signed a voluntary Code of Practice to support compliance, although the legal requirements apply regardless of participation.

The enforcement phase also introduces obligations for providers of general-purpose AI models, including documentation requirements, copyright compliance policies and publication of summaries of training data. Providers of the most advanced models considered capable of creating systemic risks must also meet additional safety and security requirements.

Responsibility for enforcement is shared among different authorities. The EU AI Office oversees general-purpose AI models and certain AI systems linked to major online platforms, while national authorities supervise most other systems. The European Data Protection Supervisor handles AI systems used by EU institutions.

Further AI Act obligations will enter into force gradually. Rules for stand-alone high-risk AI systems will apply from 2 December 2027, while high-risk systems embedded in regulated products will be covered from 2 August 2028. New prohibitions on certain harmful AI-generated content, including non-consensual intimate content and child sexual abuse material, will take effect on 2 December 2026.