Menlo Security Ships First Browser Security Platform for AI Agent Governance: Instruction-Data Separation Neutralizes Goal Hijacking

Menlo Security·medium signal

Menlo Security launched a browser security platform on March 18 designed to govern both human users and autonomous AI agents through a unified control plane. The key technical innovation is instruction-data separation ensuring agents never mistake malicious data for legitimate commands, neutralizing goal hijacking and lateral movement at the architectural level. The platform enforces least-privileged agent governance with granular controls preventing unauthorized data exfiltration, protecting 8 million users and millions of simultaneous AI agent sessions across 1,000+ enterprises including 8 of the 10 largest financial institutions.