Agent Resilience can map an agent’s memory, settings, and connected systems, then restore a trusted state. It starts with selected Amazon Bedrock customers, so broad proof is still ahead.
New to this? Read it in simple words
- On September 16, Cohesity introduced Agent Resilience, a backup tool for AI agents.
- Cohesity says it can map an agent’s memory, settings, and connected systems. After a failure, it is designed to restore a trusted version.
- An agent can change many records before a person sees the problem. So recovery must include both the agent and the work it changed.
- For now, it is only for selected Amazon Bedrock customers. So it has not yet been proven widely.
- AI agent
- An AI that takes steps on its own to finish a task, such as booking or coding.
- Backup
- A saved copy of data, kept so it can be restored after a problem.
An agent has more state than a normal application
An AI agent may depend on prompts, memory, settings, guardrails, credentials, runtimes, and connected data. A normal database backup may protect some business records. It may not restore the exact state that guided the agent.
Cohesity says Agent Topology maps these parts and their connections. The product can protect memory and configuration with saved recovery points. It can also cover databases, files, and other systems touched by the agent.
This is a practical idea. An agent can change many records before a person sees the problem. Recovery must therefore include both the agent state and the work that the agent changed.
Cohesity starts with selected Amazon Bedrock customers. Microsoft and Google support remains on the roadmap.
The first release is limited
The product starts with Amazon Bedrock AgentCore and Bedrock Agents. It is available now to selected customers. Cohesity targets general availability by the end of 2026 and lists Microsoft and Google support on its roadmap.
The company says it uses immutable backups and clean recovery environments. Those controls can reduce risk. Customers still need to test how often snapshots occur and how much recent work could be lost.
A restore is not automatically safe. Teams must confirm that the recovery point is clean. They also need to rotate exposed credentials, review connected systems, and understand why the agent failed.
Recovery tests should be routine, not emergency theatre
Cohesity reports that many surveyed organisations feel unprepared for harmful agent actions. This is company-sponsored research. It shows concern, but it does not prove that the new product solves every risk.
A useful test should start with a clear failure. Teams can corrupt agent memory, change a setting, or let a test agent alter safe sample data. They should then measure detection, restore time, lost work, and remaining damage.
Monitoring, access limits, and human approval still matter. Recovery is the last safety layer, not the first. It becomes valuable when teams practise it and record what worked before a real incident arrives.
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1Primary source · September 16, 2026Cohesity introduces Agent Resilience to protect and recover AI agent infrastructure Cohesity
- 2
- 3Primary source · September 16, 2026Cohesity Agent Resilience: protect and recover the infrastructure that powers AI agents Cohesity
We compared the launch release, product page, and technical blog. All performance and survey figures come from Cohesity. We separate features available to selected customers from roadmap items and wider future availability.