Support Telecomtalk
Highlights
- NVIDIA has launched its Open Agent Safety Platform.
- OpenShell and Sentry provide software and hardware-level controls.
- NVIDIA says Sentry can stop agents outside their boundaries in milliseconds.
Andrew Bonwick
Vice President of Product Development at Relm Insurance
Madhav Sheth
CEO of Ai+ Smartphone
Stephen Rose
CEO Render Networks
As AI agents grow more powerful and take on jobs, NVIDIA is introducing new tools to keep them safe and within set limits. The company has launched the Open Agent Safety Platform, which combines both software and hardware features. These help track, limit and shut down agents when needed.
- Make Telecom Talk My Trusted Source
Key Highlights
- NVIDIA has launched its Open Agent Safety Platform.
- OpenShell and Sentry provide software and hardware-level controls.
- NVIDIA says Sentry can stop agents outside their boundaries in milliseconds.
NVIDIA Launches Safety Platform for AI Agents
NVIDIA announced the platform on September 28. It brings together two components: NVIDIA OpenShell and NVIDIA Sentry. OpenShell is a secure runtime that creates a safety zone around an AI agent while it runs. It monitors what the agent does. It enforces rules about what the agent can touch or do. NVIDIA says this software works on its Vera CPUs but can also run on systems like those from Arm and Intel.
Sentry adds another level of protection. It runs on NVIDIA BlueField-4 data processing units. Acts as an independent observer. It monitors the agent’s behaviour at all times. If the agent tries to go beyond its allowed boundaries—like accessing files or running forbidden commands—Sentry can isolate and stop it in just milliseconds.
Also Read : Meta Muse AI App Surges in Popularity
This platform is meant for a kind of AI system one that moves beyond simple responses. AI agents can now use tools, connect with systems, and complete multi-step tasks. That independence makes security more important. Without controls, they could make mistakes or cause problems.
