AI Security
1Password Introduces Secure Agent Access Without Exposing Your Password
A new service lets AI agents log into accounts on your behalf while keeping your credentials hidden from the agent itself.
TechCrunch — AI
Password manager 1Password has rolled out 1Password for Agents, a new platform designed to solve a growing tension between the capabilities users want from AI agents and the security risks of sharing passwords.
How It Works#
The system creates a trusted communication channel between an AI agent, the user, and 1Password. When an agent needs to log into an account:
- The agent sends a unique digital identity key to 1Password
- 1Password verifies the agent's credentials
- The user receives an alert and manually confirms the login request
- 1Password then fills in the credentials on the agent's behalf, allowing it to complete its task
This approach eliminates the need for users to hand over their actual passwords to AI assistants — a practice that creates copies of sensitive data across systems users no longer control.
Growing Concern About Agent Access#
The timing of the launch reflects mounting anxiety around autonomous AI systems. On the same day 1Password announced the platform, reporting emerged that U.S. banks have begun detecting AI agent activity within customer accounts. While this isn't necessarily malicious, the problem runs deeper: most financial institutions struggle to distinguish between agentic activity initiated by users and unauthorized access.
Banking executives characterized the situation as troubling, with one Chase representative emphasizing that "customers should not give their usernames and passwords to anyone."
Supported Agents#
1Password for Agents currently works with a range of popular AI assistants, including Claude Duet (Anthropic), Dots and ChatGPT Work (OpenAI), Grok Bot, Muse (Meta), and agents from Town. The company indicates it is working to expand compatibility with additional services.
