1Password launches security tool for AI agents

1Password has introduced ‘1Password for Agents’, a new security platform designed to allow AI assistants to log into web accounts without ever seeing or storing the user’s actual passwords. As AI agents increasingly take on autonomous tasks such as booking travel, managing bills, or shopping, the risk of users sharing raw credentials in chat windows or prompt histories has grown.

Under this new system, when an AI agent encounters a login screen, it requests access using a unique digital key. The user receives a notification on their device to approve or deny the specific request. Once authorized, 1Password automatically injects the credentials directly into the website’s login form. This ensures the AI can complete the task while the password remains protected within the encrypted vault.

The service is being rolled out with initial support for Meta’s Muse and SpaceXAI’s Grok Bot, with upcoming integrations for OpenAI’s ChatGPT Work and Dots, Anthropic’s Claude Duet, and Town’s Townie. The feature is available at no additional cost to existing Individual and Family subscribers.

While the tool addresses credential exposure, some experts note it introduces new systemic risks. By granting an agent access to an active session, the agent gains operational control, which could lead to unintended actions or security breaches if the agent is compromised via prompt injection.