Back to speaking
Workshop
Technical governance of agentic AI
Workshop for more than a hundred auditors, internal auditors and AI developers.
14 April 2026 · Wolters Kluwer · for SAAF
More than a hundred auditors, internal auditors and AI developers, Wolters Kluwer.
SAAF is the framework for auditing AI agents. Using live examples, the workshop showed where the technical control of such agents falls short.
What was shown
- Direct request. A guardrail refuses a direct request for sensitive data.
- Detour. The same data is still released once the request comes in indirectly or through another agent.
- EU AI Act. This touches on the robustness requirements of Article 15(4).