Rubrik unveils Agent Identity to secure the actions of AI agents within organisations

Unveiled at the Black Hat conference, this system controls access and permissions for AI agents on a per-tool basis, with the option to reverse their actions.

Rubrik unveiled Agent Identity at the Black Hat conference, a solution designed to address an increasingly critical blind spot: most organisations currently lack the means to monitor and control, at the necessary speed and scale, the access and actions of AI agents deployed within their systems.

The solution comprises four components: real-time observability of agents and context protocols, access control per individual tool call, the enforcement of security policies at runtime, and a ‘Rewind’ function that allows an agent’s actions to be undone. Three security checks are performed sequentially before each execution: behavioural analysis, infrastructure-level access control, and identity verification using short-lived tokens. The system integrates with Okta and Microsoft Entra ID.