Autodesk adopts Permiso Security for AI-agent monitoring

Per reporting by SiliconANGLE and Dealroom, Permiso Security launched new AI agent runtime security capabilities and Autodesk signed on as the launch customer. SiliconANGLE reports Autodesk is deploying the capabilities to secure agents operating across its products, global workforce and cloud infrastructure. Dealroom and SiliconANGLE describe the release as offering agent discovery, runtime identity attribution, tool and infrastructure observability, runtime detection of anomalous behaviour, skill sandboxing and identity-first controls.

According to Dealroom, the platform provides continuous monitoring of agent runs, tool calls and data access and ties every agent action to a specific identity at runtime. Dealroom reports the architecture is agentless and API-based and includes machine-speed kill switches. SiliconANGLE reports Permiso highlights visibility across agents, sub-agents and Model Context Protocol servers and the infrastructure those agents run on.

SiliconANGLE quotes Sebastian Goodwin, chief trust officer at Autodesk: "Autodesk is investing significantly in AI across our workforce, infrastructure and products." SiliconANGLE also quotes Permiso co-founder and co-CEO Jason Martin: "You are putting a deterministic capability on a non-deterministic brain. Agents will do things they were not supposed to do."

Companies deploying autonomous or semi-autonomous AI agents often lose visibility after authentication; public coverage frames Permiso's emphasis on runtime attribution as addressing that gap rather than posture-only solutions. For practitioners, tying actions to runtime identities and instrumenting tool calls and data access reduces forensic blind spots and shortens mean-time-to-detect for agent-driven incidents.

Observed patterns in comparable transitions show enterprises accelerating agent usage across product features and internal workflows, increasing the need for tooling that can correlate identity, tool invocation and data access in real time. Identity-first controls and machine-speed remediation features are becoming a common set of requirements among vendors targeting agent security.

Track adoption signals beyond a launch-customer announcement: integrations with major cloud providers, Model Context Protocol observability plugins, and third-party validation of runtime detection efficacy. Also watch for case studies showing how runtime attribution changes incident response timelines versus posture-based controls.