1Password Unveils Tool Letting AI Agents Sign In Without Seeing Users' Passwords

Password manager 1Password has introduced a tool designed to let artificial intelligence agents sign in to personal accounts on a user's behalf without ever handling the user's actual credentials. The launch, announced Thursday, comes as concern grows over how much access consumers are handing to automated assistants.

The product, called 1Password for Agents, works by setting up a line of communication among the agent, the user and the password manager. When an agent tries to log in for someone, it sends a unique digital key to 1Password so the service can verify the agent's digital identity. The user then receives an alert about the attempt and can confirm it personally.

Once the user approves, 1Password signs the agent in by entering the credentials itself. That allows the agent to proceed with whatever task it was assigned, without the password passing through the agent at any point.

Users can pair the service with the agent of their choice. The options named include Anthropic's Claude Duet, OpenAI's Dots and ChatGPT Work, Grok Bot, Meta's Muse, and agents from Town, an AI assistant startup. The company says it is working with other companies to broaden the list of supported agents.

In a blog post, 1Password described the risk it is trying to address. "Handing over a password, secret, or credential is a common shortcut to an agent, but can leave people with sensitive information in places they can no longer control as a copy gets saved for the next time," the company said. It added that its platform gives people "a trusted way to delegate tasks to personal AI agents while keeping control of the passwords, secrets, and credentials those tasks require."

The launch coincided with a report from Axios that U.S. banks have started detecting agentic activity inside customer accounts. According to the report, this does not necessarily point to rogue agents stealing from accounts. Instead, it reflects consumers increasingly letting automated services act for them, including on financial matters.

Even so, the potential for problems remains considerable. The central difficulty appears to be that most banks cannot determine whether a given piece of agentic activity was actually requested by the account holder.

Bankers interviewed by Axios described the situation as "horrifying." A Chase spokesperson also told the outlet that customers "should not give their usernames and passwords to anyone."