From 2 August 2026 the European Commission's Artificial Intelligence Office (AI Office), together with national authorities, will start monitoring the application of the EU AI Act, the European Commission announced.
One immediate effect of the regulation is stricter transparency requirements. From that date certain AI systems must inform users when they are interacting with artificial intelligence rather than a human. They must also disclose when an image, video or audio recording was created or modified by AI — meaning that deepfake images, videos and audio must be labeled.
According to the Commission, these measures aim to reduce deception and manipulation and to help people make better-informed decisions when they encounter online content or interactive systems.
Who is affected?
- The rules apply immediately to chatbots and other interactive AI systems that communicate directly with users; these systems must clearly state that the interlocutor is AI.
- Deepfake images, videos and audio recordings must be labeled in line with the transparency requirements.
Obligations for general-purpose model developers and additional duties
The regulation can now be enforced against developers of general-purpose AI models. These developers face additional obligations to manage risks, in particular those related to:
- chemical, biological, radiological and nuclear (CBRN) incidents,
- risks arising from loss of human control,
- cybercrime and other harmful manipulation,
- impacts that threaten fundamental rights.
The aim is to ensure that development of large-capacity, general-purpose systems includes measures to prevent and mitigate potential harms.
Implementation and national responsibilities
The European Commission urged member states to designate competent national authorities to enforce the regulation and to provide them with necessary resources. The Commission also published an initial list of more than 180 organisations that have signed a practical code on transparency for AI-generated content.
The Commission plans for the AI Office and national authorities to cooperate in monitoring compliance and taking action in cases of non-compliance.
Why this matters
Mandatory labeling and increased transparency are intended to curb online manipulation and strengthen public trust in AI-generated content. For developers and service providers the new rules represent both legal and practical obligations, particularly for systems with broad capabilities and high potential impact.
