[key points]

Key points

• •Controlling agentic systems involves precisely defining identity and limiting agent capabilities, treating them as users with clearly defined roles and permissions.
• •It is essential to control the tools used by agents by approving and limiting their access, managing the tool supply chain, and prohibiting automatic tool-chaining without explicit policy.
• •Implementing these measures is crucial for cyber-resilience and preventing misuse, in accordance with regulations such as the EU AI Act.