EU prepares new AI Act oversight rules

Starting August 2, 2026, the European Union will begin the operational application of the AI Act, placing oversight with the European Commission’s AI Office in cooperation with national authorities. New transparency obligations will require chatbots and other conversational systems to tell users they are interacting with an automated system, while deepfakes and AI-generated or altered media must be clearly identified and marked in machine-readable form.

The rules also impose additional duties on advanced models that may create systemic risks, including risks linked to chemical, biological, radiological and nuclear accidents, cyberattacks, malicious manipulation, loss of control and threats to fundamental rights. Certain practices will be prohibited, including systems that manipulate behavior, exploit vulnerabilities or make discriminatory or unjustified assessments affecting fundamental rights.

Supervision will be divided among the AI Office, competent national authorities and the European Data Protection Supervisor. Scientific support will come from a Scientific Panel composed of sixty independent experts, alongside Professor Alessandro Abate of the University of Oxford as the AI Office’s Lead Scientific Adviser.

The AI Office has created reporting and whistleblowing channels for alleged violations. Under the amended timetable, high-risk AI provisions take effect on December 2, 2027, rules for high-risk systems in regulated products apply from August 2, 2028, and new prohibitions on non-consensual sexually explicit content and child sexual abuse material take effect on December 2, 2026.