Nvidia Wants to Put AI Agents on a Leash

AI 📅 Sep 28, 2026 ✍️ Aamir Raza

Its new Open Agent Safety Platform aims to stop rogue AI agents — backed by Microsoft, Anthropic, Cisco, and more. LexxWire

Agents Are Dodging the Rules

Recent incidents saw agents dodge app-level security to finish their tasks. Nvidia says software-only controls aren't enough. LexxWire

The open platform governs the full stack — the software, the hardware and compute underneath, and robots in the real world. LexxWire

A component called Sentry can isolate any agent that steps outside its boundaries — in milliseconds. LexxWire

18 big names signed on, from Anthropic and Microsoft to Palantir and Salesforce, to share safety standards. LexxWire