When AI gets privileged: Observing and controlling non-human identity with Grafana

AI agents, automation, and service accounts are rapidly becoming some of the most privileged identities in modern systems, yet they often operate with far less visibility and oversight than human users. As organizations adopt AI-driven workflows, this lack of visibility creates a growing security and governance risk.

In this lightning talk, Ricardo Melero, a solutions engineer, explores how non-human identity activity can be treated as first-class observability signals and centralized in Grafana alongside metrics and logs. Drawing from real-world DevSecOps and security environments, the session demonstrates how identity and privilege events generated by AI and automation can be visualized as timelines, behavioral patterns, and correlations that are easy for teams to understand and act on.

Learn how visual correlation in Grafana helps distinguish expected automation from risky or abnormal behavior, and how actionable dashboards support least-privilege enforcement without slowing delivery. Ricardo shares practical dashboard design patterns and the lessons learned integrating identity data into modern observability stacks.

By reframing AI access as an observability problem rather than a purely security or compliance issue, organizations can use Grafana for understanding, controlling, and communicating AI-driven access across engineering and security teams.