Cohesity adds recovery capabilities for AI agents and the data they manage

Cohesity has introduced Cohesity Agent Resilience. This new Cohesity Data Cloud capability will discover, protect, and recover the infrastructure behind enterprise AI agents.

A unified view of an agent and the state it depends on. (Source: Cohesity)

The company outlined its vision for Autonomous Cyber Resilience, which uses agentic workflows to automate its five-step cyber resilience framework: protect data, identity, applications, and agents; help ensure recoverability in all scenarios; remediate cyber and AI threats; practice application recovery; and optimize data and AI risk posture.

Cohesity also introduced the Cohesity AI Resilience Academy, beginning with a free course that helps organizations manage risk as AI systems move from advising people to taking action. Cohesity announced the new capability, vision, and academy today at Cohesity Catalyst.

Key highlights

Across the enterprise, AI agents are querying databases, updating workflows, and using privileged credentials, often without a person reviewing each action. Gartner predicts up to 40% of enterprise applications will include integrated task-specific agents by 2026, up from less than 5% today. Governance and observability tools can identify unexpected behavior, but they cannot restore data or systems after an agent has changed, corrupted, or deleted them. Cohesity Agent Resilience is designed to address that recovery gap.

New research findings from the fifth annual Cohesity Global Cyber Resilience Report further underscore the need to strengthen AI infrastructure resilience. Fifty-six percent of organizations said they were not well prepared to detect or contain unintended actions by AI agents and automated workflows. At the same time, 58% were not very confident in their ability to verify the integrity of AI models and related data following a cyberattack.

“Detection can tell you an AI agent went off course. It cannot undo the changes,” said 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.”

Cohesity Agent Resilience protects the infrastructure behind AI agents

An AI agent’s behavior depends on several components, including agent memory, configuration, credentials, guardrails, and workflows. At launch, Cohesity Agent Resilience protects agent memory and configuration using the same snapshot architecture, immutable backups, and clean-room recovery that Cohesity customers already rely on to protect sensitive workloads and data across on-premises, cloud, and SaaS environments.

Cohesity Agent Resilience includes two core functions:

• Protect agent state: Protect agent memory and configuration with point-in-time recovery capabilities that help teams restore agents to a known-good state following memory corruption, misconfiguration, or malicious activity.
• Protect what agents manage: Protect databases, file systems, and other services agents interact with, enabling precise recovery of affected resources when needed.

Agent topology provides teams with a unified view of each agent, including its memory stores, connected applications, databases, and supporting infrastructure. That makes it easier to understand agent dependencies, assess protection coverage, and identify the resources required to restore trusted operations during incidents.

“As AI agents take on more business-critical work, we need to think beyond monitoring and make sure we can recover when something goes wrong,” said Srikanth Kuntamukkala, CIS Global Lead for Partner Sales, Cognizant. “Being able to restore an agent’s state, and the data it may have changed, could give our teams a more practical way to manage that risk as we move agentic applications into production.”

At launch, Agent Resilience integrates with Amazon Bedrock. Support for additional agent platforms, including offerings from Microsoft and Google, is on the roadmap.

Cohesity Agent Resilience is available now to select customers, with general availability targeted for the end of the year.

A path toward autonomous cyber resilience

Cohesity also outlined how agentic workflows could help operationalize its five-step cyber resilience framework. The aim is to move beyond manually configured, point-in-time plans toward a system that continuously discovers and protects data, validates recovery readiness, and adapts response and recovery plans as conditions change, with a human in the loop. As attackers increasingly use AI to move from initial compromise to operational disruption, resilience plans that rely on manual triage and sequential approvals may not keep pace.

From stated objectives to automated execution

In this model, teams would define objectives through Cohesity Copilot rather than manually configuring policies across applications. For example, administrators would be able to specify that all applications and data required to sustain critical business operations during a cyber incident must meet specified recovery time objectives (RTOs) and recovery point objectives (RPOs) and adhere to specified threat scanning and recovery testing requirements. Cohesity Data Cloud would then assess the resilience posture of those workloads, recommend appropriate protection policies, threat-scanning strategies, and recovery-rehearsal plans, and present those recommendations for approval before execution.

During a cyber incident, administrators can initiate automated response and recovery workflows to assess the scope of impact, identify indicators of attacker activity, and prepare an isolated recovery environment to accelerate investigation and recovery efforts.

Automated protection for newly discovered sensitive data

As a step toward that model, Cohesity introduced enhancements to existing capabilities. Customers with Cohesity Data Cloud Enterprise Edition and Cohesity DSPM can now automate the protection of newly discovered sensitive data that is not already protected. Cohesity continuously discovers and classifies sensitive data, evaluates its protection status, and can trigger protection actions based on policy. Administrators can configure and manage these workflows directly or through Cohesity Copilot.

Built on Cohesity RecoveryAgent, open to customers’ AI tools

The Autonomous Cyber Resilience model builds on Cohesity RecoveryAgent, which orchestrates key parts of incident response and recovery and helps security and IT teams work from the same plan. Cohesity Maestro, with expanded capabilities expected later in 2026, will connect Cohesity’s protection, response, and recovery capabilities to customers’ preferred AI tools, including Claude, ChatGPT, and Gemini, as well as Cohesity Helios, the company’s unified management console.

Cohesity plans to add more automation as the Autonomous Cyber Resilience vision moves toward commercial availability.

“Cyber resilience is not a one-time assessment. It requires teams to protect data, test recovery, detect threats, rehearse responses, and continually improve their risk posture,” said Murthy. “Agentic workflows can take on more of that routine work and help keep the plan current.”

Cohesity launches AI Resilience Academy

The Cohesity AI Resilience Academy is a new learning path within Cohesity Academy. Its first course, Foundations of AI Resilience with Cohesity, starts with a simple distinction: AI assistants make recommendations; AI agents can make changes. The free, self-paced course takes 25 to 30 minutes and focuses on concepts and vocabulary rather than product configuration.

After completing the course, learners will be able to:

• Define an AI agent and distinguish it from an AI assistant.
• Explain the additional risks introduced when AI moves from advising to acting.
• Identify two gaps in the AI lifecycle, one before an agent acts and one after.
• Describe, at a high level, how Cohesity Gaia Catalog addresses the upstream gap.
• Describe, at a high level, how Cohesity Agent Resilience addresses the downstream gap.

Cohesity Gaia Catalog and Agent Resilience are included to illustrate Cohesity’s broader architecture and product direction. Some capabilities discussed or shown may not yet be generally available.

Catalyst attendees will receive a direct enrollment link in the post-event follow-up and can begin the course immediately. The course will remain exclusive to attendees for approximately two weeks before appearing in the broader Cohesity Academy catalog.

Those who complete the course will earn a Credly badge titled “Foundations of AI Resilience with Cohesity,” which they can share on LinkedIn. The course will remain a free entry point into Cohesity Academy. Cohesity may offer more advanced AI Resilience courses for a fee over time.