DigiCert has added support for NVIDIA’s Open Agent Safety Platform, beginning with the OpenShell runtime, so organizations can attach verifiable identities and limited authority to autonomous AI agents. The integration combines DigiCert’s AI Trust Manager with a runtime that controls an agent’s access to files, networks, credentials and other resources.
OpenShell enforces policy outside the agent’s process and records decisions to allow or block actions. DigiCert’s contribution is an AI Passport, a cryptographically signed credential linking an agent to an accountable owner. The company says other systems can verify the credential without sharing the same identity provider, an approach designed for agents that interact across company boundaries.
Policy-based permissions, which DigiCert calls visas, can specify the systems, data and actions an agent may use and how long that authority lasts. When an agent attempts a prohibited action, an automated control can block it. Depending on policies set by the owner or credential issuer, DigiCert says the same control can revoke the agent’s authority and quarantine it.
OpenShell determines whether a particular action is permitted within its environment, while the AI Passport supplies an independently verifiable statement about which agent is acting and who authorized it. DigiCert says agents, models and Model Context Protocol servers can also be signed and associated with a bill of materials before execution. Model Context Protocol is a framework through which agents connect to external tools and data. Those connections are among the interactions that organizations may need to authenticate and restrict.
Other identity vendors have begun testing credentials for cross-company AI-agent transactions, while Omada has added runtime controls for agent access through its acquisition of EmpowerID. DigiCert’s support for OpenShell brings those two concerns together: the agent’s identity and the permissions enforced as it operates.
DigiCert AI Trust Manager currently supports OpenShell as the first component of NVIDIA’s new platform. NVIDIA’s platform is intended to secure agents from testing through deployment, with OpenShell providing the initial policy boundary. AI Trust Manager became generally available earlier this month. The DigiCert integration extends its existing credential and revocation functions to the OpenShell environment, where policies are enforced outside each agent’s process.
–
By the ID Tech Editorial Team
