1Password limits AI agent access to single tasks

1Password limits AI agent access to single tasks
1Password is treating AI agents as hybrid identities that blend human and machine characteristics, complicating attribution in audit logs. The company rejects standing privilege for agents, instead granting just-in-time access scoped to individual tasks. Its Credential Broker releases secrets only when needed, keeping them away from the agent and underlying model. CTO Nancy Wang compared the approach to verifying an intern's work before allowing them to proceed. 1Password and Okta are also backing shared identity standards so verified agent authorization can carry across both platforms.
Read the original article →