Agent Resilience targets a new recovery gap as autonomous AI systems gain access to enterprise data, applications and privileged resources
As enterprise AI agents move from recommending actions to executing them, organisations face a new dimension of cyber risk: what happens when an autonomous system changes, corrupts or deletes critical data and infrastructure? Cohesity is addressing that gap with Cohesity Agent Resilience, a new capability designed to discover, protect and recover the infrastructure supporting AI agents.
Introduced at Cohesity Catalyst, the capability initially supports Amazon Bedrock AgentCore and Amazon Bedrock Agents, with Microsoft and Google agent platforms planned for the roadmap. It is available to select customers, with general availability targeted for the end of 2026.
The launch reflects a broader change in the AI risk equation. Traditional observability and governance tools can identify unexpected agent behaviour, but detection alone does not restore affected systems. Cohesity’s approach extends its existing snapshot architecture, immutable backups and clean-room recovery capabilities to AI-agent infrastructure.
“Detection can tell you an AI agent went off course. It cannot undo the changes.”
— Vasu Murthy, Chief Product Officer, Cohesity
“Cohesity Agent Resilience provides a precise recovery path for both the agents and the resources they impact. Our vision for Autonomous Cyber Resilience is to use agentic workflows to reduce manual work from response and recovery,” said Murthy.
At launch, the platform protects agent memory and configuration, while also protecting databases, file systems and other services that agents interact with. Agent topology provides a consolidated view of an agent’s memory stores, connected applications, databases and supporting infrastructure, helping teams understand dependencies and protection coverage.
The need is underscored by Cohesity’s Global Cyber Resilience Report, which found that 56% of organisations surveyed were not well prepared to detect or contain unintended actions by AI agents and automated workflows. Another 58% lacked strong confidence in their ability to verify the integrity of AI models and associated data following a cyberattack.
Cohesity is also outlining an Autonomous Cyber Resilience vision that uses agentic workflows to automate elements of protection, threat remediation, recovery testing and incident response, while retaining human approval.
The company has further introduced automated protection for newly discovered sensitive data through Cohesity Data Cloud Enterprise Edition and Cohesity DSPM.
Alongside the technology launch, Cohesity is introducing its AI Resilience Academy, beginning with a free, self-paced course designed to help organisations understand the additional risks that emerge when AI systems move from advising to acting.
The broader message is clear: as AI agents gain operational authority, resilience must extend beyond protecting the data they consume to protecting the agents themselves—and the systems they can change.
