Where does AI agent security fail?

A filter that catches 99.9% of prompt injections still lets 10 through at 10,000 requests a day. What happens next depends on the systems the agent can reach.

In this video, Kayla Williams explains the failure points in AI agent security: from designs falling short, through controls that measure the wrong thing, to human gates that turn into rubber-stamping at scale.

Learn where the inherent weaknesses lie, so you can be better prepared to secure the agents in your organization.