1Password Launches Secure Logins for AI Agents

This is the right security boundary for agentic software: delegate access without handing over the secret. But protecting the password does not automatically limit what an agent can do after signing in.

• –Digital-key identity and per-request approval give users a clearer control point than pasting credentials into prompts.
• –The model never sees the password or one-time code, reducing exposure through context windows, logs, and agent memory.
• –Once authenticated, the agent can still read data or act within the account’s existing permissions.
• –Developers will need consent flows, scoped accounts, audit trails, and stronger post-login controls as agents move into financial and workplace systems.
• –Availability remains limited to supported agent integrations rather than every AI tool.