1Password ties AI agent access to individual tasks

The problem 1Password is describing is not that agents are new, but that they blur a line identity systems were built around. An agent logs in, carries credentials and acts on someone's behalf, so it has characteristics of both a human and a machine. That overlap is what makes the audit trail ambiguous: an agent may appear as the person it works for even though software took the action. Nancy Wang's answer is to stop treating the agent as a fixed account and instead verify it task by task, the way you would check an intern's work before letting them move to the next project.

That thinking has already been packaged into a product. 1Password rejects standing privilege for humans, machines and agents alike, granting access just in time and checking why it is needed, and it recently turned that into a privileged access product scoped to a single task. The company's Credential Broker releases secrets only when needed, without exposing them to the agent or the underlying model.

The piece that is still in motion is portability. 1Password and Okta are backing shared identity standards so an agent's verified identity and authorization context can carry across their systems. For security teams, the practical question is whether that context survives the trip between vendors; if it does not, the distinction between who authorized an action and what performed it may remain hard to reconstruct.