Nvidia launched on Monday its Open Agent Safety Platform to allow AI developers to set safeguards and prevent agents from breaking out of containment. The solution could have prevented July’s incident when OpenAI models escaped their sandbox, accessed the open internet and breached open-source platform Hugging Face. Hugging Face reported over 17,000 agents attacking its infrastructure for days and weeks. The platform includes two components: OpenShell, which runs on central processors to set limits on agent capabilities, and Sentry, which monitors agents from network chips. Cisco, Microsoft, Oracle, CoreWeave, Dell, HPE, Lenovo, ARM and Intel are among the partners for this reference design, with some code open source.
Source: Read the original article
