EU AI Act rules begin applying to general-purpose models

European Union Artificial Intelligence Act obligations for general-purpose AI models began applying on 2 August 2025. Providers placing new models on the EU market became subject to technical-documentation, information-sharing, copyright-policy and training-content-summary requirements.

Models classified as posing systemic risk faced additional duties for evaluation, adversarial testing, risk assessment, incident reporting and cybersecurity. Classification could follow a computing threshold or Commission designation based on capability and impact, not simply a brand's popularity.

Existing models received a later compliance period, and enforcement structures continued to develop. A voluntary code of practice could help providers demonstrate compliance but did not replace the regulation or make signatories immune from investigation.

General-purpose models can be integrated into many downstream systems. The provider's obligations are therefore different from duties of a deployer using a model in a high-risk medical, employment or public-service application. Compliance must be mapped to role and use.

The AI Act timetable establishes the milestone. This News item distinguishes these rules from prohibited practices that applied in February and high-risk obligations scheduled later. It records the first operational framework for foundation-model providers without claiming that every technical standard or enforcement question was already settled.

Providers outside the EU can fall within scope when offering models in the union, making the regime relevant beyond European companies. Territorial reach depends on market placement and use, not the nationality of a model's developers alone.