Xage Security Enhances Zero Trust for Artificial Intelligence

Xage Security introduced enhancements to its Zero Trust for Artificial Intelligence (AI) platform. The company said the update added deterministic visibility and precise control over agent behavior, aimed at helping organizations govern AI agents when moving them toward production deployments.

Xage described increased activity around AI agents as organizations connected agents to APIs, SaaS platforms, databases, internal applications, cloud services, and operational technology (OT) environments. The release also said users deploy “shadow AI” agents that can receive broad access to critical resources, while organizations often lack deterministic visibility and controls to govern what agents can see, do, and change.

According to Xage, its new Zero Trust capabilities cover the full AI interaction chain, including users, agents, LLMs, tools, and cloud or internal applications. The platform combines two capabilities: Xage Agent Sentry, which encapsulates the agent wherever it runs and monitors inputs and outputs, and Xage Resource Gateway, which sits in front of critical resources to govern interactions. The release added that the approach controls actions at network-interaction, local event, and OS-call levels, and it produces detailed logs for governance and audit.

As part of its described use cases, Xage said it could block compromised agents from exfiltrating data or damaging core organizational resources, citing an OpenClaw agent demonstration where an agent was hacked and manipulated. The release also stated that Xage assigns secure digital identities upon onboarding, defines agent-specific policies by role, resource, and time-bound need, and detects unmanaged “shadow AI” agents for either onboarding or removal. It further said the system blocks harmful actions if an agent is compromised and supports anomaly detection through behavioral baselining, early warning systems, and SIEM/SOC integration. “AI is ready to move beyond the sandbox, but organizations cannot safely deploy it in production unless they know exactly what agents are doing and can control the actions they take,” said Duncan Greatwood, CEO of Xage Security.