Rubrik ties AI security tools to AWS Bedrock AgentCore

Rubrik has announced an integration between Rubrik Agent Cloud and Amazon Bedrock AgentCore, extending its AI security tools to customers building agents on Amazon Web Services.

The integration is aimed at organisations using Amazon Bedrock AgentCore to build and manage AI agents at scale. Rubrik said its software will add semantic governance and operational safeguards by feeding policy signals into AgentCore.

Under the arrangement, AgentCore's policy engine will enforce controls across agent traffic, including exchanges between agents and tools, agents and large language models, and agent-to-agent communications. This will allow security teams to apply Rubrik's detection tools alongside automated enforcement at the gateway, outside the agent's reasoning loop, the company said.

The setup is designed to combine probabilistic detection with deterministic enforcement, allowing actions taken by an AI agent to be allowed or denied consistently. That addresses a key concern for businesses adopting autonomous software agents, where the ability to monitor intent and block harmful actions remains limited.

Key tools

Customers using Amazon Bedrock AgentCore will be able to access several functions from Rubrik Agent Cloud, including SAGE, its AI governance engine for monitoring and controlling autonomous agents in real time.

They will also be able to use Agent Inventory, which is intended to discover agents running on the Amazon Bedrock AgentCore runtime and provide visibility into risk, access permissions and policy violations. Another feature, Agent Rewind, is designed to reverse destructive actions carried out by an autonomous agent.

Administrators will also be able to manage AI security policies through a unified control interface covering AWS and hybrid cloud data environments. Existing Rubrik Agent Cloud customers will be able to connect their policies to AgentCore and apply SAGE language-based policies to block risky actions.

The announcement comes as technology suppliers and their customers try to put more structure around the use of AI agents in business systems. These agents are increasingly being developed to act on behalf of users, interact with software tools and data sources, and make limited operational decisions without direct human intervention.

That has brought greater scrutiny to governance and recovery tools. Businesses are looking not only for ways to inspect what agents are doing, but also for mechanisms to stop harmful behaviour quickly and undo actions when systems fail or operate outside expected boundaries.

AWS has positioned Amazon Bedrock AgentCore as infrastructure for building, connecting and managing agents. Rubrik's integration adds its control layer to that environment, with a focus on policy enforcement and visibility into agent behaviour.

Devvret Rishi, General Manager, AI, Rubrik, said the company sees a growing need for stronger governance as organisations deploy more autonomous systems. "Speed has always been the enterprise mandate, but today, safety is the prerequisite," Rishi said.

He added: "Cybersecurity leaders estimate that in the next year, half or more of attacks they face will be driven by agentic AI. Rubrik's upcoming integration with AWS Bedrock AgentCore gives enterprises the confidence to deploy AI agents quickly with trustworthy governance."

The integration is Rubrik's latest step in expanding its AI security products for AWS users. The company has been building out Rubrik Agent Cloud as a platform for monitoring, auditing and controlling agentic actions across enterprise environments.

Rubrik is listed on the New York Stock Exchange under the ticker RBRK. Its broader business spans data protection, cyber resilience and AI-related operations for cloud and hybrid infrastructure.