European Commission's power to enforce the AI Act against general-purpose model providers activates August 2 — one year after the obligations themselves took effect
The AI Act's obligations for general-purpose AI model providers took effect on 2 August 2025. The Commission's power to actually enforce them against providers activates on 2 August 2026. For twelve months the rules existed without a regulator able to act on them; in two days that gap closes. Providers shipping under free and open-source licences with public weights are excused from two of the four core obligations but still owe a copyright compliance policy and a training-data summary.
The substantive piece is the enforcement gap closing, not the rules changing. The obligations have been law for a year — what arrives on 2 August is the Commission's standing to act against a provider that ignored them. Every general-purpose model shipped into the EU market since August 2025 has been accumulating compliance exposure against a regulator that could not yet move. That asymmetry ends this week, and it ends retroactively in the sense that the obligations were never suspended.
The open-source carve-out is narrower than the shorthand suggests. Providers releasing under a genuinely free and open licence with publicly available weights are excused from technical documentation and downstream-provider information — two of four. They still owe a copyright compliance policy and a published training-data summary. That second requirement is the sharp one: a training-data summary is precisely the disclosure most labs, open-weight included, have been least willing to make.
The competitive read for H2 2026 procurement is that EU-market compliance stops being a roadmap item and becomes a gating question in vendor selection. Enterprises buying model access into European operations now have a regulator to point at, and a date. Expect procurement questionnaires to grow a training-data-summary line within the quarter.
ComplexDiscovery — Open weights, open questions: the letter that redrew the AI policy fight → · International Center for Law & Economics — Open Models, Closed Minds: AI Policy Keeps Regulating the Wrong Thing →