Commvault has launched a new software tool, AI Protect, designed to monitor AI agents within AWS, Azure, and GCP environments and roll back unauthorized actions.
The software aims to provide a safety net for enterprises deploying autonomous agents that may behave unpredictably or access sensitive data without authorization.
According to go.theregister.com, the product is part of a broader push by the 30-year-old company to establish itself within the growing AI resilience market.
Protecting the AI stack
Commvault’s field CTO, Vidya Shankaran, told the outlet that enterprises are currently racing to deploy AI agents while struggling to maintain proper governance.
She noted that many organizations overlook the necessity of protecting vector databases, which serve as the essential memory for large language models.
"The vector database stores the embeddings that large language models rely on. If it’s compromised or lost, you would either have to rebuild it from scratch or retrain the model. No one has that luxury of time anymore," Shankaran said.
AI Protect works by establishing a baseline of normal behavior for agents across cloud platforms. It flags deviations, such as an agent suddenly attempting to access payroll information it previously could not reach.
When an anomaly is detected, the system notifies administrators and provides options to roll out original configuration files or repair corrupted data.
Shankaran clarified that the tool is not designed to directly stop or control third-party agents.
"We would rather stay in our own swim lane," she told the outlet, "and not really overreach and say, ‘Hey, Salesforce agent, don’t do that.’"
Commvault also introduced Data Activate and AI Studio alongside the new security tool. Data Activate allows organizations to use existing backup copies to train AI models, reducing the computational strain on live production systems.