Rubrik launches Agent Cloud to govern enterprise AI

Rubrik has launched Rubrik Agent Cloud, a product designed to monitor, govern, and reverse the actions of AI agents operating inside corporate systems.

Rubrik is pitching the release as a response to rising spending on agentic AI projects and growing concern about how these systems behave once connected to business applications and data.

Rubrik Agent Cloud sits between applications, AI agents, and the large language models behind them. It focuses on visibility into which agents exist, what they access, and what they do. It also adds policy controls over prompts and responses, plus a recovery feature that links agent actions to known-good snapshots.

Production risk

Agentic AI systems often run with broad permissions and can act quickly across multiple tools. Rubrik says that combination increases operational risk when models make errors or produce incorrect outputs.

"While building an agent might take only days or weeks, deploying it often takes months. I've found that the number one blocker to AI adoption isn't the quality of the models - it is the lack of a consistent framework for governance and guardrails in managing AI risk," said Devvret Rishi, General Manager of AI at Rubrik.

Rubrik describes agents as large language models with access to tools and non-human identities that can touch large volumes of enterprise data. It pointed to the potential impact of mistakes in software and data operations, including accidental deletion of critical information.

"These types of mistakes can lead to some reticence among the leadership tasked with AI oversight. And good AI governance is a necessary part of successful enterprise AI deployments. But currently AI governance often lives only on paper, leading to 'death by committee' where deployments are stalled by manual reviews. Suddenly, your impressive new AI agent is obsolete before it is even deployed," Rishi said.

Three pillars

The product groups its functions into three areas: continuous monitoring and observability, dynamic governance, and remediation.

For monitoring, Rubrik Agent Cloud scans an environment and builds what it calls a dynamic agentic inventory. The inventory is meant to show which agents are running, when they were created, and how they are used, along with the tools and data they access.

Governance controls apply to both inputs and outputs, including prompts sent to agents and the responses and tool calls they generate. The product includes predefined policies such as personally identifiable information detection. It also supports custom policies written in natural language, enforced on interactions using small language models.

The remediation feature, called Rubrik Agent Rewind, ties into Rubrik's existing cyber resilience and backup capabilities. It correlates an agent's actions with a previous healthy snapshot.

The goal is to make it possible to undo changes when an agent takes an unwanted action, such as deleting data. Rubrik says it can restore assets from backup, including a dropped database or deleted Salesforce opportunities.

Compatibility approach

Rubrik says Rubrik Agent Cloud is vendor-neutral and works across different agent frameworks and deployment models.

Custom-built agents can integrate through a standalone AI gateway, including agents built on open-source frameworks such as LangChain, or through direct calls to models such as OpenAI and Claude. For low-code tools, Rubrik points to backend integrations for Microsoft Copilot Studio. It also supports discovery and deployment for agents running on endpoints.

The launch moves Rubrik more directly into the operational control layer for enterprise AI, an area that has expanded quickly as companies shift from experimentation to broader rollouts. Governance and oversight have become central issues as agent-based systems gain access to customer records, internal documents, developer tools, and automated workflows.

"The goal of Rubrik Agent Cloud is to move governance from a theoretical bottleneck to a real-time enabler. By baking guardrails directly into the infrastructure, we give teams the safety they need to innovate and deploy at speed," said Rishi.