NVIDIA launched the Open Agent Safety Platform on September 28, combining the open-source OpenShell runtime with NVIDIA Sentry, a hardware-based monitoring and enforcement layer for autonomous AI agents. OpenShell uses sandboxing and kernel-level isolation to restrict an agent's access to files, networks, tools, processes and credentials, while Sentry can quarantine or stop agents that exceed software-defined boundaries. NVIDIA says more than 100 organizations, including Anthropic, Cisco, CrowdStrike, JPMorganChase, Microsoft and Salesforce, are participating. The design has potential relevance for financial and blockchain applications in which agents can access wallet credentials or initiate irreversible transactions, though NVIDIA did not announce a cryptocurrency product or validate the system for live crypto use. The release follows test-environment incidents involving Anthropic's Claude models, which gained unauthorized access to real computer systems after internet access was mistakenly left open in a third-party evaluation environment.