What happened
On August 2, 2026, the European Commission's AI Office and national authorities gained applicable enforcement powers under the EU AI Act. The Act had received the Council's final approval on May 21, 2024; this later commencement made specified powers and duties operative rather than adopting a second law. Commission commencement account.
The AI Office can require information, obtain access to general-purpose models for evaluation, seek corrective measures and impose penalties through Commission decisions. The framework also supplies complaint and whistleblower channels. Transparency requirements tell people when they are interacting with AI and require specified synthetic-content marking or labeling. Scope and exceptions matter: this is not a statement that every provision of the Act is already applicable. The Commission's current account places Annex III high-risk-system requirements in December 2027 and requirements for high-risk AI embedded in regulated products in August 2028. Enforcement powers and phased timetable.
The institutional change gives public authorities compulsory access and corrective leverage over major suppliers, rather than depending entirely on voluntary disclosure. It also makes some AI mediation more visible to people and downstream institutions. The significance comes from enforceable authority and avenues for challenge, not from regulatory strictness alone. Effective staffing, accessible remedies and actual enforcement remain practical tests.
Potential direction of power
Leaning dispersion · Dispersion +2
Enforceable information, evaluation-access and corrective powers give public authorities leverage over AI suppliers beyond voluntary cooperation.
Window of contestability
Opens
Public enforcement and complaint channels expand opportunities to challenge covered AI practices.
