1Password rolls out tool letting AI agents log in without seeing passwords

1Password introduced 1Password for Agents on Thursday, a feature that lets AI assistants sign into a user’s accounts without ever exposing the user’s login details. When an assistant tries to access an account, a unique digital token is sent to 1Password to verify the assistant’s identity. The user receives an alert and must approve the request before the manager fills in the credentials on the assistant’s behalf. The service supports a range of agents, including Claude Duet from Anthropic, OpenAI’s Dots and ChatGPT Work, Grok Bot, Meta’s Muse, and bots from the startup Town. 1Password says it is partnering with more providers to broaden the offering. In a blog post, the company warned that handing over passwords to agents is a common shortcut but can leave sensitive data in places the user can no longer control. It positions the new tool as a trusted way to delegate tasks while keeping full control over secrets and credentials. The launch coincided with an Axios report that U.S. banks are now spotting AI‑driven activity inside customer accounts. Bank officials described the trend as “horrifying,” and a Chase spokesperson told Axios that customers should never share usernames or passwords with anyone.

Context

AI assistants are being used for everyday tasks, from booking rides to managing finances, which raises new security challenges. Password managers traditionally store credentials; this new layer adds user‑approved mediation, giving users a way to benefit from AI convenience without surrendering full access.