1Password Launches Secure Agentic Autofill with Human-in-the-Loop to Protect Credentials from AI Agents

1Password has announced a Secure Agentic Autofill system designed to protect credentials from AI agents by inserting a human-in-the-loop into the autofill process. The approach requires explicit human approval before any credentials are transmitted, with the approval captured via a pop-up dialog and authentication methods such as Touch ID or a PIN.

Once approved, an end-to-end encrypted channel is established between the browser extension and the web interface to securely inject the credentials, ensuring the AI agent never gains direct access to usernames or passwords. The system is described as a safeguard against AI agents that could otherwise retain or autofill credentials after unlocking a password manager.

Currently, Secure Agentic Autofill is available for early access through Browserbase, a platform for AI agent browser environments and tools, allowing developers to test the feature and balance automation with oversight. According to 1Password, the trade-off is a temporary interruption to autonomous workflows in exchange for enhanced security.