This is very cool: NVIDIA launched an open source security layer for AI agents, potentially THE security layer. The basic idea: don’t trust the AI to police itself. Agents run in a sandbox that controls what they can access, while a separate system watches what they do and can shut them down if they go rogue. OpenShell is free and open source. The biz model feels a bit like the CUDA playbook: make your software layer a standard, then make NVIDIA hardware the best place to run it. They…