1Password is addressing the security challenges posed by AI agents, which function as hybrid identities between human and machine users. CTO Nancy Wang noted that these agents complicate identity controls because they can act on behalf of individuals, making it difficult for security teams to attribute actions in audit logs.
The company has introduced a privileged access product that restricts agents to task-based, just-in-time permissions. This approach ensures that agents must verify the completion of a specific task with appropriate authorization before moving on to subsequent operations, mirroring verification processes used for human interns.
1Password is collaborating with Okta to support shared identity standards, allowing authorization contexts to persist across systems. The company’s Credential Broker manages this by releasing secrets only when necessary, preventing them from being exposed directly to the AI model or the agent itself.
Looking ahead, 1Password anticipates a shift where applications function as thin clients, with code execution moving to remote sandboxes. The company emphasizes that secure access management must occur within cloud environments to maintain integrity as AI agents become more prevalent in enterprise workflows.
