Commvault Partners with Zscaler to Govern and Recover Enterprise AI Agents

Commvault announced a strategic partnership with Zscaler on November 10, 2026, combining AI-use discovery and policy enforcement with recovery for data, applications and identities. The companies plan to help enterprises govern production AI agents and roll back harmful or unintended activity.

The partnership addresses the operational gap between controlling an agent and repairing what it changes. Zscaler Security for AI identifies enterprise use of AI applications and enforces policies for users and AI agents. Commvault’s AI and cyber-resilience technology restores data, applications and identities after attacks, disruptions or outages, creating a proposed path from detection through recovery.

That division matters when experimental agents become production systems with direct access to business data. Security and IT teams must determine which AI tools employees and developers use, what data those tools share, which activities to permit or block, and whether unwanted changes can be reversed quickly.

Planned Discovery, Governance and Recovery

The companies named AI discovery, runtime governance, and agent detection and recovery as use cases for the partnership. The announcement presents these as planned collaboration areas rather than a generally available joint product.

One workstream would expand enterprise-wide discovery and maintain continuous visibility into active AI applications, models and tools. Another would create guardrails that flag unsanctioned AI activity and prevent sensitive-data exposure. The third pairs real-time protection for autonomous agent operations with rapid, clean recovery following a security event or risky AI action.

AgentRecover Supplies the Recovery Model

Alongside the partnership, Commvault provided a more detailed view of its recovery approach in a separate AgentRecover announcement. The new capabilities support Amazon Bedrock AgentCore and Microsoft Foundry, platforms that provide services for building, deploying and operating production agents. AgentRecover is designed to audit agent activity, restore agents and recover affected data and systems across environments.

Commvault Chief Technology and AI Officer Pranay Ahlawat said agents do not behave deterministically and that a single agent decision can trigger changes across data, applications and infrastructure on multiple platforms. Such an unintended action can propagate at machine speed, affecting data integrity, application availability and business operations.

For supported platforms, AgentRecover is designed to discover agents and map the data, applications, identities, configurations and infrastructure with which they interact. That dependency map is intended to reveal protection gaps. Its auditing functions are designed to establish what an agent did and what it may have affected across platforms, helping teams prioritize recovery work.

Recovery is designed to cover both the agent and the affected resources while concentrating on what the agent changed. Customers can also use AgentRecover with Commvault’s identity recovery and change auditing to examine agents’ actions, the data they changed and the identities through which they acted.

Commvault is targeting early access for Microsoft Foundry in November 2026 and for Amazon Bedrock AgentCore in December 2026. Those targets apply to AgentRecover’s rollout; the Zscaler announcement does not make them availability dates for the partnership’s planned work.

The partnership connects preventive controls with a recovery path for changes that slip past them. Its immediate significance is the division of labor: Zscaler would provide visibility and enforcement, while Commvault would address restoration after an incident. Turning those planned use cases into working discovery, governance and agent-recovery capabilities for joint customers is the central execution test.