Nvidia's OpenShell safety system opens to all, adds Sentry watchdog
Nvidia's OpenShell agent safety system, first announced in March, is now available to everyone, limiting what agents can do, with a new Sentry watchdog that stops agents trying to cross those limits.




This is actually a good thing, finally there will be some limitation on AI agents 🔒 Nvidia's OpenShell safety system which was first announced in March is now available for everyone, it puts clear limits on what AI agents can do, and Nvidia also added a new watchdog called Sentry that can stop any agent trying to cross those limits and lock it away, honestly this was much needed because in the last few months some AI agents hacked into companies and even poked around US and Australian government websites