25 February 2027 14:00 - 14:30
Governing Agentic AI Safety & Security
As AI agents start acting autonomously across enterprise systems, the attack surface moves from the model to the runtime, where prompt injection, tool misuse and data leakage play out in real time.
Drawing on experience building AI security at a global bank, this keynote breaks down what governing agentic AI looks like in practice: runtime guardrails, red-teaming agentic systems, security gates for onboarding models, and metrics such as attack success rate that show leadership whether controls actually work.
Expect a practical blueprint for scaling agentic AI without leaving security as an afterthought.
25 February 2027 16:30 - 17:00
Panel | Governing agentic AI: Finding the balance between autonomy, trust, and risk
AI agents can now take actions across enterprise systems without waiting for approval, and the governance models built for predictive models and chatbots werenāt designed for that.
Panelists from security, AI governance and engineering share how they decide how much autonomy an agent gets, where human approval should sit, and which controls make autonomy safe to scale, including permissions, runtime guardrails, audit trails and red-teaming.