Commvault has unveiled AI Protect, a system that gives enterprises an “undo” button for autonomous artificial intelligence agents operating across cloud platforms like AWS, Azure, and Google Cloud.
The new governance tool tackles a mounting challenge in enterprise AI deployment: agents that can execute destructive actions faster than human oversight can prevent them. These AI systems routinely perform complex sequences of operations—deleting databases, altering security configurations, or modifying access controls—while attempting to solve assigned problems. Traditional permission-based security fails because agents can combine individually approved actions in unexpected ways, creating cascading effects that human administrators struggle to predict or contain.
AI Protect monitors these digital workers continuously, tracking their API calls and data interactions across cloud infrastructure in real time. When an agent goes astray, the system can precisely reverse its specific actions while leaving human-generated changes intact. This granular rollback capability addresses what Commvault’s Chief Technology and AI Officer Pranay Ahlawat describes as agents’ tendency to “mutate state across data, systems, and configurations” in ways that compound rapidly and become difficult to trace.
The system also discovers AI agents that may be operating invisibly within enterprise infrastructure—a growing blind spot as organizations deploy more autonomous tools. Rather than requiring a complete system restoration to a previous backup, AI Protect can surgically undo problematic agent behavior while maintaining legitimate work performed during the same timeframe.
This development signals the maturation of AI governance from theoretical frameworks to practical tools that address the operational realities of autonomous systems in enterprise environments.
rtificialintelligence-news.com ↗