The gist: From August 2026, the EU Commission will have full control functions over AI providers. GPAI providers are subject to procedural and substantive obligations, with a one-year transition period. Various actors – market surveillance authorities, downstream providers and scientific bodies – can support enforcement.
As of 2 August 2026, European authorities can exercise their control functions over providers of general-purpose AI models. While the obligations have been in force since August 2025, providers receive a one-year transition period before the European Commission may exercise its enforcement powers.
The European Commission has extensive powers to monitor and enforce the AI Act against providers of general-purpose AI models (GPAI). These powers include the right to request documentation and information, conduct assessments, order measures to ensure compliance and mitigate risks, and impose fines.
Providers of GPAI models must fulfil both procedural obligations – including cooperation with the AI Office – as well as substantive requirements concerning the development and documentation of the models.
In addition to the Commission, other actors play an important role in enforcing the legislation: national market surveillance authorities can call on the Commission to exercise its powers, downstream providers can file complaints against GPAI providers, and a scientific committee can point the AI Office to systemic or concrete risks.
Providers of models published before 2 August 2025 must be fully compliant by 2 August 2027.
Source: artificialintelligenceact.eu