Astrix Security Expands AI Agent Platform With Full-Stack Discovery and Policy Enforcement

New updates have been reported about Astrix Security.

Astrix Security has significantly expanded its AI agent security platform, unveiling at RSA Conference 2026 a multi-layer discovery and control system designed to surface both sanctioned and shadow AI agents, MCP servers, and non-human identities across the enterprise. The platform addresses the growing gap between rapid AI agent deployment and slower governance processes, which often leave agents in production with access to critical systems but no formal review or enforcement guardrails.

At the core of the update is a four-pronged discovery architecture that integrates directly with leading AI platforms, fingerprints non-human identities such as OAuth apps and service accounts, ingests telemetry from EDR and network sensors, and supports proprietary services via a bring-your-own-service model. These feeds are consolidated into the Astrix Platform, which maps each agent to its credentials, reachable resources, accountable owner, and automatically calculated risk score, enabling prioritized remediation based on blast radius.

The company has also extended its Agent Control Plane with a real-time policy engine called Agent Policies that lets security teams define granular allow, flag, and block rules by user, department, platform, and resource type, with a default rule that flags unrecognized shadow AI activity before actions execute. By moving from mere visibility to active enforcement, Astrix positions its platform as a de facto control layer for governing agent actions, aiming to reduce the risk of unsupervised agents that can now read, write, delete, and execute across core systems.

Co-founder and president Idan Gour framed the launch as essential to managing the emerging AI attack surface, arguing that complete agent inventory and runtime controls are prerequisites for scaling AI productivity programs safely and deciding which agents to standardize and roll out enterprise-wide. For customers and prospects, the expanded capabilities could influence vendor selection, internal AI governance models, and budget allocation toward identity-centric controls for AI agents and other non-human identities, an area where Astrix claims demand is accelerating as NHIs vastly outnumber human users.