Agentic AI governance gap identified as core EU AI Act compliance challenge

With the EU AI Act’s high-risk requirements set to become broadly enforceable from 2 August 2026, a growing body of analysis published in April 2026 identifies agentic AI governance as the most difficult compliance challenge. Agentic systems that act across multiple tools, APIs and data sources without continuous human confirmation struggle to meet Article 9 risk management and Article 13 transparency requirements because their decision trails are difficult to reconstruct after the fact. Organisations without a registry of every agent, its granted permissions and its audit log cannot demonstrate compliance to national market surveillance authorities. The issue is acute for law enforcement deployers: high-risk AI systems used in policing, border control and criminal justice fall under Annex III and carry maximum fines of €35 million or 7% of global turnover. Only 8 of 27 EU member states had designated their national competent authorities by the August 2025 deadline, creating a governance gap ahead of the August 2026 enforcement date.