News

Rubrik Hits 'Undo' on AI Agent Mistakes

• By David Ramel
• 08/12/2025

Rubrik is giving enterprises a way to hit "undo" when AI agents go rogue. The company this week introduced Agent Rewind, powered by recently acquired Predibase AI infrastructure, which tracks, audits, and allows rollback of AI agent actions. The capability is designed to address what Rubrik describes as a growing need for visibility and control as organizations deploy autonomous agents into production systems.

A company representative told Virtualization & Cloud Review that Agent Rewind was developed to address a fundamental blind spot within agent activity. While AI agents have significant potential, they're prone to mistakes that can disrupt businesses, just like humans. Current observability tools can detail what happened, but they can't explain why it happened or how to reverse high-risk AI actions. "It's like having a security camera that records a break-in but can't help you restore what was stolen," the rep said.

Rubrik said Agent Rewind records every action taken by an AI agent, creates an audit trail linking the action back to the originating prompt, and enables organizations to restore affected data and configurations to a safe recovery point. According to the company's announcement, the offering will integrate with platforms such as Agentforce, Microsoft Copilot Studio, and Amazon Bedrock Agents, as well as custom-built agents.

AI Agents and Non-Human Error
Rubrik cited a recent study finding that AI agents "are frequently becoming disoriented, choosing incorrect shortcuts, and struggling to complete even simple multi-step tasks." The company pointed to incidents where agents made unintended record updates, broke workflows, or deleted production data, with issues often going unnoticed until they caused operational or business disruption.

Johnny Yu, research manager at IDC, said, "Agentic AI introduces the concept of 'non-human error,' and as with its human counterpart, organizations should explore solutions that allow them to correct potentially catastrophic mistakes made by agentic AI."

Examples in a Rubrik blog post include a quoting agent severing the link between a deal and a customer account without alert, or an onboarding automation revoking access for hundreds of active users due to a flawed trigger. "These aren't bugs. They're high-blast-radius agent actions with real operational fallout," the blog noted.

Key Features

• Context-Enriched Visibility -- Surfaces agent behavior, tool use, and impact while mapping each action back to its root cause, including prompts and execution plans.
• Auditability -- Traces agent actions to the specific data or applications accessed, identifying high-risk activities.
• Safe Rollback -- Uses Rubrik Security Cloud to rewind changes to files, databases, configurations, or repositories.
• Broad Compatibility -- Works with a range of agent frameworks and custom AI agents.

Anneka Gupta, chief product officer at Rubrik, said, "Agent Rewind integrates Predibase's advanced AI infrastructure with Rubrik's recovery capabilities to enable enterprises to embrace agentic AI confidently. Today's organizations will now have a clear process to trace, audit, and safely rewind undesired AI actions."

AI Resilience
The company positioned Agent Rewind as moving beyond cyber resilience toward what it calls AI resilience, describing the need to let agents operate with broad access and autonomy while maintaining the ability to reverse problematic actions. The blog states, "The answer isn't to lock your AI agents down. It's to build the guardrails that let them move fast without surrendering your control."

Chad Pallett, chief information security officer at BioIVT, said, "When using AI, there is a need for observability and secure rollback. In a market craving true observability and remediation, Agent Rewind is the answer I've been waiting for."

A demo is available along with an educational video and a solutions brief for more information. To apply for the early access waitlist, go here.