NVIDIA Watch posted an update
Nvidia has unveiled OpenShell, a reference architecture for building, containing and observing AI agents. Jensen Huang says it uses a zero-trust approach, with agents given only the access and tools they need, and can intervene if one tries to go rogue.
Why it mattersHuang told Yahoo Finance that Nvidia is working with more than 100 companies on agent-management features using OpenShell. He also said a Hugging Face exercise found its website would not have been attacked if OpenShell had been in place. That is Huang’s account of the exercise, not an independently established result. The useful idea is less “trust the agent” and more “limit what it can do”.
Discuss: Should agent makers make restricted access the default, even if it makes agents less convenient?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.