1Password Addresses Critical AI Browser Agent Security Gap

1Password has unveiled a new integration to address security gaps in AI-enabled browser agents, aiming to prevent credential leakage as organisations increasingly deploy agentised browsers. The feature, named Secure Agentic Autofill, is a collaboration with Browserbase and is designed to let agents authenticate in the browser without exposing credentials, while keeping a human in the loop.

According to Nancy Wang, SVP of engineering at 1Password, Secure Agentic Autofill helps solve visibility, control, and governance concerns as AI agents automate tasks in the browser, such as form filling or travel bookings. The approach relies on delivering credentials just-in-time through an end-to-end encrypted channel and autofilling only after a human approval, leveraging 1Password’s zero-knowledge security model.

However, security experts warn that traditional identity protections have not kept pace with AI, noting risks from prompt injection and the potential for credentials to be exposed if agents are tricked into visiting malicious sites. The article, by Arielle Waldman for Dark Reading, was published on 10 October 2025. according to Dark Reading.