As reported by Bloomberg, on September 28 (local time), NVIDIA rolled out an open agent safety platform characterized by a dual-tier architecture. By integrating OpenShell software with Sentry hardware monitoring, it constructs a dual-defense mechanism for AI agents, adept at thwarting any out-of-bounds actions within milliseconds. Such a platform could have averted the July incident in which OpenAI managed to infiltrate Hugging Face. Presently, over 100 organizations have embraced relevant technologies, covering a wide range of sectors including AI models, enterprise software, finance, energy, and robotics.
