I think this is a step in the right direction. However, there seems to be too many encumbrances and too much reliance of existing technology to be sufficient to address AI agent safety.

We at ArmorIQ welcome it and think there is more to it then meets the eye. https://armoriq.substack.com/p/nvidia-is-moving-agent-safety...