Cohesity launches Agent Resilience to protect and recover AI agent data, debuting with Amazon Bedrock integration

Cohesity introduced Agent Resilience within Cohesity Data Cloud at Cohesity Catalyst, adding discovery, snapshot-based protection, immutable backups, and clean-room recovery for AI agent memory, configuration, and agent-managed databases and file systems. It…

At Cohesity Catalyst, Cohesity announced Agent Resilience, a capability within Cohesity Data Cloud that discovers, protects, and recovers the infrastructure behind enterprise AI agents. It protects agent state (memory and configuration) and agent-managed resources such as databases and file systems using snapshot architecture, immutable backups, clean-room recovery, and point-in-time recovery, while an agent topology view gives a unified picture of memory stores, connected applications, and databases. The offering debuts with Amazon Bedrock integration; support for Microsoft and Google platforms is planned, and general availability is targeted for year-end 2026. Cohesity cited Gartner's prediction that up to 40% of enterprise applications will include task-specific agents by 2026, and company research finding that 56% of organizations are unprepared to detect or contain unintended agent actions while 58% lack confidence in verifying AI model integrity after attacks. The company also outlined an Autonomous Cyber Resilience vision using agentic workflows, with Cohesity Copilot automating RTO/RPO policy recommendations, and introduced the AI Resilience Academy. The launch was also featured in Help Net Security's September 18, 2026 weekly roundup of security product releases.

• Agent Resilience was announced at Cohesity Catalyst as part of Cohesity Data Cloud (reported 2026-09-16).
• It discovers, protects, and recovers AI agent memory, configuration, and agent-managed data, including databases and file systems.
• Protection uses snapshot architecture, immutable backups, clean-room recovery, and point-in-time recovery for agent state and agent-managed resources.
• Agent topology provides a unified view of memory stores, connected applications, and databases.
• Launch integration is with Amazon Bedrock; Microsoft and Google platform support is planned.
• General availability is targeted for year-end 2026.
• Gartner predicts up to 40% of enterprise applications will include task-specific agents by 2026 (cited by Cohesity).
• Cohesity research: 56% of organizations are unprepared to detect or contain unintended AI agent actions; 58% lack confidence in verifying AI model integrity after attacks.