“While building an agent might take only days or weeks, deploying it often takes months,” Devvret Rishi (pictured), general manager of AI at Rubrik, said in a statement.

“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.”

AI agents, typically powered by large language models, can access enterprise systems, execute tasks and interact with sensitive data at machine speed. But their non-deterministic nature means they can make errors or unintended changes, raising the risk of operational disruption or data loss.

Rubrik Agent Cloud sits between enterprise applications, AI agents, and their underlying models, providing visibility into agent activity and enforcing policies on how agents access data and perform actions.

A key component of the platform is Rubrik Agent Rewind, which allows organisations to recover quickly from agent-driven errors by restoring systems and data from previous backups.

Rubrik said the platform also enables organisations to define governance policies using natural language, allowing guardrails to be enforced dynamically without requiring complex manual configuration.

“The goal of Rubrik Agent Cloud is to move governance from a theoretical bottleneck to a real-time enabler,” Rishi said.

“By baking guardrails directly into the infrastructure, we give teams the safety they need to innovate and deploy at speed.”

The platform is designed to work across a range of AI development environments, including open-source frameworks, direct integrations with major AI models and enterprise AI platforms such as Microsoft Copilot Studio.