1Password will let you hand your data security over to your preferred AI agent – Engadget
After introducing Claude integration in July, 1Password just launched a platform for other major AI companies including Muse from Meta, OpenAI’s Dots and SpaceXAI Grok Bot. The new 1Password for Agents gives those models access to your logins so they can complete tasks for you, without ever receiving your username, password or vault contents.
1Password is pitching the feature as a way to make your personal AI assistant dreams come true for tasks like booking restaurants, buying gifts and planning vacations. Doing those things usually requires login credentials, but given AI agents’ penchant for going rogue, doing so can be fraught. That’s why 1Password came up with a login system that ensures the models can’t see them.
“An agent should be able to do the job it’s been asked to do without getting permanent access to everything behind the account, and 1Password provides the trust layer that makes this possible across personal agent experiences,” said 1Passwords chief product officer, Aashish Tripathi.
To do that, 1Password for Agents lets agents use stored credentials from your 1Password vault without them reaching the model. When agents reach a password check in the course of performing actions you requested, they use a special key to verify their identity with 1Password. 1Password then notifies you about which agent needs access and why, and fills in the required information (not the agent) if you approve.
Access is granted per session, limited to specific, approved items and not carried over. Since passwords and one-time MFA codes are injected through a secure channel managed by 1Password, they stay outside the model.
The company said that 1Password now works with Muse, Claude Duet, ChatGPT Work and Dots, Grok Bot and Town. For each of those, it provides approved access when needed and lets users approve sign-ins. Passwords, usernames and MFA codes remain outside each agent’s view for each, 1Password says, so they can’t spread beyond your own devices.
Still, common sense is required when giving agents unfettered access to accounts. Though they won’t spill your password secrets, they can certainly do things like make product ordering errors or book the wrong vacation with no option to cancel. As such, you’d want to be extremely careful if deploying one for such things. The company has yet to reveal other details like pricing or availability, but we should hear more about those things soon.
