AI agents are increasing taking on sensitive tasks in organizations, including reading emails, raising tickets, moving money and calling other agents with credentials their owners issued and rarely revisit. An IDC report found that 55% of IT leaders already run agentic AI in production.

Despite this extensive activity, organizations are relying on security controls designed to control people and software. Controls like SIEM and IAM are unable to adequately manage the access and actions of dynamic and persistent AI agents.

This webinar will delve into this issue, highlighting what agentic AI means in operational terms, where the risk is growing fastest, and which controls you can deploy to close the gap.

The session will firstly offer a look on the NCSC's August 2026 guidance on managing agentic AI cyber risk and recent autonomous intrusions against critical infrastructure. An expert panel will then look at how the guidance fares against the realities of accreditation, procurement and live incident response, offering best practice advice for security teams to take on the agentic AI challenge.

Sign up to learn:

• How to tell an AI agent from an automation or a chatbot, and why that distinction changes your threat model and your chain of accountability
• Where agentic risk is growing fastest, including delegated credentials, tool and MCP sprawl, prompt and response injection
• What enforcement at the point of action looks like in practice, covering agent identity, per-action authority, revocation and evidential audit
• The questions you should put to any vendor selling AI agents