Rubrik unveils Agent Cloud to govern & remediate AI agent risks

Rubrik has announced the launch of Rubrik Agent Cloud, a platform designed to facilitate the adoption, monitoring and auditing of AI agents within enterprises while addressing operational risks associated with Agentic AI.

The new solution aims to provide organisations with tools to oversee agentic actions, enforce controls, fine-tune agent accuracy, and crucially, allow remediation of mistakes made by AI agents. Rubrik Agent Cloud is described as the first solution in the sector to deliver monitoring, governance and remediation for Agentic AI from a single source, and is positioned to help organisations adopt agents at scale.

AI agent risks and oversight

As businesses accelerate efforts in AI transformation, many are facing new challenges linked to the deployment of Agentic AI. Such agents, which include automated systems capable of making decisions and performing tasks without direct human input, present significant opportunities but also substantial operational risks. These risks include AI hallucinations, inadvertent deletions, and the potential for rapid propagation of mistakes or unauthorised actions. In light of these issues, Rubrik has framed its new offering as a response to mounting concerns about the speed and scale at which damage could occur.

Bipul Sinha, CEO, Chairman, and Co-Founder of Rubrik, commented on the dilemma confronting IT and security professionals as they manage enterprise AI deployments.

"IT and security leaders often don't know what their AI agents are doing or how to undo their mistakes. Rubrik wants to help them answer: 'What agents do I have?' 'What are they capable of doing?' 'How are they performing?' 'What did they do?' and 'Can I undo that when they screw up?' AI agents have the potential to cause 10x the damage in 1/10 of the time. With Rubrik Agent Cloud, we uniquely address this challenge by leveraging our leadership in data, identity, and resilience to help our customers deploy AI agents with peace of mind."

Agent lifecycle management

Rubrik Agent Cloud incorporates a comprehensive range of features spanning the full lifecycle of AI agents, from discovery and observability through to governance and remediation. The solution is integrated within the Rubrik Platform, which brings together data, identity and application contexts to support AI operations with added security and efficiency.

The Agent Monitor component of the platform is designed to automatically discover both infrastructure and platform-based agents running across environments such as Azure, AWS, M365, OpenAI, Microsoft Copilot Studio and Amazon Bedrock. It provides continuous oversight of agent activities and data access, while maintaining immutable audit trails with extensive contextual information.

Agent Govern is introduced to help organisations track agent usage, evaluate agent performance, and control or prevent undesirable actions. It offers real-time policy enforcement on agent behaviour and actions, and integrates with enterprise identity systems to support compliant and secure operations.

Agent Remediate, which includes a feature called Agent Rewind, allows users to roll back undesirable or destructive actions initiated by agents. This provides a method to instantly undo changes without incurring downtime or data loss, enhancing the protection of critical enterprise data and maintaining system resilience.

Deployment and availability

Rubrik Agent Cloud is currently available through a limited early access programme for selected customers. The company has indicated that not all functionalities are available at this stage, with planned extensions to its feature set anticipated as part of its product development roadmap.

The solution is pitched to address a core challenge for enterprise IT leaders: maintaining control over rapidly expanding AI agent deployments, while ensuring security and compliance. Rubrik Agent Cloud's launch comes amid a broader industry conversation around trust and risk in AI operations, with stakeholders seeking tools that can both amplify operational efficiency and provide mechanisms to recover from errors introduced by AI-driven automation.