Nvidia Tames Rogue AI Agents with Open Source

🔥 What happened Nvidia launched the Open Agent Safety Platform to stop rogue AI agents. Over 100 companies – including Microsoft, JPMorgan, and Perplexity – are already using it. Nvidia claims it could have prevented the Hugging Face breach by OpenAI agents. 💡 Why it matters OpenShell formally verifies an agent's permissions, while Sentry monitors it on-chip and can quarantine it in milliseconds. This is the first serious attempt to treat agent safety as an engineering problem, not a philosophical debate. In a breach, milliseconds matter. ⚡ Our take Jensen Huang is right: safety is an engineering problem. But building an open tool that runs on rival hardware is also a brilliant move to undermine – sorry, extend – Nvidia's CUDA moat.