Red Flag? OpenAI's Agentic ChatGPT Work Signs Into Your Accounts Without You
OpenAI’s August 25 release notes added an agentic browser to ChatGPT Work that can continue a task on a login-protected site after the user signs in. If a site supports authentication, ChatGPT shows the login screen so the user can enter credentials or a security code, then the agent can keep working and may remain signed in for later tasks. OpenAI says password managers are supported, and usernames/passwords are not visible, stored, or used for training. The main tradeoff is convenience versus security: once signed in, the agent effectively has the same access as the user until the browser session is cleared manually. This creates a persistent account foothold, not just a one-time login. The feature is available in ChatGPT Work’s browser on web and mobile, with per-site session clearing in Settings > Cloud browser.
