1Password says it can fix login security for AI browser agents

1Password’s browser extension automatically fills your passwords as you browse, and the company has now created a similar tool for AI bots browsing the web on your behalf, but for a very different reason.

AI tools and browsers based on Claude, Gemini, and ChatGPT are increasingly using AI agents to browse the web, book tickets, and create Spotify playlists for you, and unlike the risk of forgetting a one-time password, an AI bot is likely to remember it and cause a breach later. 1Password’s fix for this potential risk, a new secure agent autofill feature that “injects credentials directly into the browser if, and only if, the human approves access.”

With this tool, when a browser AI agent determines that it needs login credentials, “the agent notifies 1Password that a credential is requested,” 1Password says. “At this point, 1Password identifies the appropriate credentials and requests user approval through a human workflow.” To approve a request, a human authenticates the request with something like Touch ID on their Mac, and 1Password’s tool uses an “end-to-end encrypted channel” between its extension in the browser operated by the AI ​​agent and the approving device to enter the credentials. The AI ​​and LLM agent, according to 1Password, never see the actual credentials.

Initially, Secure Agentic Autofill is available today in early access through Browserbase, which creates a browser and tools specifically designed for AI agents.