The European Commission's AI Office published an update on 9 April 2026 detailing progress under the AI Continent Action Plan, including enforcement milestones tied to the EU AI Act [source]. The announcement confirms that member states have begun formal compliance assessments of general-purpose AI models and high-risk systems deployed in the Union.

According to the update, the AI Office has received over 140 notifications from providers of general-purpose AI models since the Act's application date. The Commission states that initial assessments have identified "several cases requiring further investigation" into transparency obligations and systemic risk management, though no providers are named [source].

The document notes that national market surveillance authorities in France, Germany, and the Netherlands have opened proceedings related to alleged non-compliance with documentation and conformity assessment requirements. The AI Office states these cases involve both foundation models and downstream high-risk applications in employment and credit scoring [source].

The update also reports that the AI Office has issued guidance on prohibited practices under Article 5, including emotional recognition in workplace and educational settings. The guidance clarifies that certain biometric categorisation systems remain lawful under specific conditions, a point that has drawn criticism from civil society groups cited in the announcement [source].

The Commission emphasises that enforcement actions are ongoing and that providers have the right to respond before any penalties are imposed. The AI Office plans to publish a public register of high-risk AI systems by the end of Q2 2026, as required under the Act [source]. No specific models or providers are identified in the current announcement.