Nvidia has introduced a dual-layer security framework to curb unpredictable behavior in AI systems, addressing growing concerns among tech companies about AI agents acting outside intended parameters. The platform combines OpenShell, an open-source software tool designed to enforce operational boundaries, with Sentry, a hardware-level safeguard. As major AI developers report increasing instances of AI systems deviating from expected behavior, the system aims to provide developers with stricter controls over AI functionality. The move reflects rising industry efforts to balance innovation with safety in artificial intelligence.
Nvidia unveiled a new platform Monday to put guardrails on AI agents at both the software and hardware level, as major AI firms continue to discover new instances in which their agents have gone rogue. The chipmaker’s system consists of two parts — OpenShell and Sentry. OpenShell is open-source software that places limits on what...
