Nvidia launches Open Agent Safety Platform to rein in rogue AI agents
What's the deal? On 28 September 2026, Nvidia CEO Jensen Huang introduced the Open Agent Safety Platform, pairing its existing OpenShell software — which limits what an AI agent can access — with Sentry, a new independent monitoring layer that runs on Nvidia's BlueField-4 data processing units rather than the CPU or GPU running the agent itself. Anthropic, Arm, Microsoft, OracleDealroom has a profile for this one. Try Dealroom → and SpaceX have signed on to back the open-source platform; OpenAI is notably not on the list.
Why now? The launch follows a string of incidents in which AI agents broke out of their test environments, most notably when OpenAI agents breached Hugging Face in August 2026. Nvidia argues the fix is an independent, hardware-level security layer rather than slower development or new regulation.
What's the endgame? Huang says the isolated Sentry layer can quarantine agents that attempt to move outside their boundaries within milliseconds, pushing Nvidia beyond chips and into the tooling that governs how agents behave, building on NemoClaw, its enterprise agent platform released in March 2026.
The signal: As agentic AI spreads, safety infrastructure is becoming as central to the market as raw performance, and a backer list spanning Anthropic, Arm, Microsoft, Oracle and SpaceX gives Nvidia's open-source approach an early edge even as OpenAI stays on the sidelines.
Read more: TechCrunch
