Rubrik Tackles AI Agent Security Challenge

According to a Zero Labs study of over 1,600 respondents, 86% of IT and cybersecurity leaders believe AI agents will surpass their organization's protection capabilities within the next twelve months, while only 23% report having complete visibility into active agents in their environment. In this context, data security specialist Rubrik announces Agent Identity, a solution designed to manage access and permissions for AI agents. Integrated into Agent Cloud, the technology aims to provide a control framework for agents capable of interacting with information systems and executing actions autonomously. Dev Rishi, General Manager of AI at Rubrik, explains that agents no longer just synthesize information but act on behalf of employees, using access models originally designed for humans, and static credentials were never intended for autonomous actors. The solution leverages existing identity platforms, with integrations announced with Okta and Microsoft Entra ID. Agent Identity gives AI agents their own identity, comparable to that of human users or applications, to determine what actions an agent is authorized to perform and under what conditions. It can identify and monitor AI agents and servers using the Model Context Protocol (MCP), grant temporary permissions via just-in-time (JIT) access mechanisms, control tool calls made by agents, enforce security policies before action execution, and track activities. Technically, it secures each MCP tool call by analyzing requests through the SAGE framework and checking them against enterprise security policies, then authenticates the agent identity using temporary access tokens. If an action violates rules, it can be blocked before execution. The platform also includes Agent Rewind, a capability to undo certain actions performed by autonomous agents, addressing concerns of security teams about operational consequences of errors or unexpected AI agent behavior. According to the Zero Labs study, 88% of IT and cybersecurity leaders worry about meeting their recovery time objectives (RTO) in the face of agentic AI threats.