1Password allows Claude to log you into websites without having to view your passwords.

      TL;DR: 1Password for Claude enables the AI agent to access your logins through biometric approval, ensuring that credentials never reach the model or Anthropic’s systems.

      1Password has introduced a browser integration that allows Anthropic’s Claude to use stored credentials for web tasks without the passwords ever being disclosed to the AI model, as detailed in a blog post released on Thursday. This setup is referred to as a zero-exposure architecture: when Claude needs to log in, 1Password indicates which credential is required and its purpose, then awaits biometric approval before directly entering the login information into the webpage. Claude does not view the vault item, password, or one-time code, and access is terminated once the task is done.

      This integration addresses a key issue in agentic AI. While browser-based agents like Claude can browse websites, complete forms, and make purchases, users have traditionally had to either share their passwords or manage these tasks themselves when they reach a login page. According to 1Password, this is the first browser integration that enables an agent to use credentials without granting them direct access.

      After autofill, 1Password verifies whether any secrets were exposed on the page. If the submission fails, the extension clears the entered values before resuming control for Claude. Throughout the process, the credential remains encrypted and under 1Password's control.

      The latest updates from the EU tech scene include insights from our founder Boris and some dubious AI-generated art. Subscribe to receive this weekly round-up for free!

      The launch also brings Agentic Mode, a feature in the 1Password browser extension that automatically secures the vault when a compatible AI agent takes over. The agent can only access logins and one-time codes specifically authorized for the current task, while the rest of the vault remains inaccessible. Agentic Mode activates even if the 1Password-Claude integration is not set up and is compatible with agents beyond Claude.

      The timing of this release is significant, given that security researchers recently showcased how AI browsers could be manipulated into revealing user credentials via prompt injection attacks, including Anthropic’s Claude extension. 1Password CTO Nancy Wang stated in the announcement that the solution is not to disclose secrets to agents but to allow users to authorize an agent to use a credential without showing it to the agent. She emphasized that this distinction is central to establishing trust in AI agents.

      1Password for Claude is currently available on Mac for business, family, and individual plans, requiring the 1Password desktop app, browser extension, Claude desktop app, and Claude browser extension. The company, which has recently acquired Israeli startup Apono to manage AI agent access within enterprise systems, plans to extend support for payment cards and identity information post-launch.

      CNET’s password manager expert Joe Supan expressed that he would typically be cautious about granting an AI agent access to his password manager; however, he noted that 1Password seems to have implemented several effective safeguards, especially biometric authentication for each login. This integration represents the first instance of a major password manager establishing a secure channel for an AI agent to utilize credentials in real time, rather than exposing them to the model's context. Yet, it remains to be seen whether this method can withstand the types of prompt injection attacks that have previously affected AI browsers.