Permiso launches AI agent runtime security

Per Silicon Angle, Permiso Security Inc. launched AI agent runtime security capabilities on Thursday designed to provide security teams with continuous visibility into agent activity across both cloud and on-premises environments. Autodesk Inc. has been announced as the launch customer for these new features.

The new capabilities extend Permiso’s unified identity platform to address the growing challenge of securing AI agents, which are increasingly making autonomous decisions and interacting with various systems without direct human oversight. Security teams can now discover all agents, track their runs, events, tool calls, and data access across different environments. Autodesk is implementing these features to enhance the security of agents operating within its products, global workforce, and cloud infrastructure.

The platform offers six key capabilities, including agent discovery, runtime identity attribution, observability into touched resources, detection of over-privileged access and policy violations, behavioral sandboxing, and identity-first controls. These features are informed by Permiso's threat research, which has identified vulnerabilities like LLMjacking and malicious agent skills.

Source: Silicon Angle