ChatGPT Work Signs Into Websites for Users Without Seeing Passwords
The authentication feature moves AI agents from the public web to the authenticated web, where most account-based tasks occur. Rapid user growth amplifies the scale at which agents operate across private accounts and services.
Codex and ChatGPT Work users grew from roughly 6 million to 25 million in about seven weeks, with the platform adding approximately 5 million users every 10 days; growth inflected on July 9, coinciding with the launches of GPT-5.6 and ChatGPT Work. The platform's browser authentication feature lets the agent operate on login-gated websites without handling user credentials: when a site requires sign-in, the agent pauses, lets the user authenticate manually inside a virtual browser, then resumes once control is returned. Screenshots are disabled during that takeover step, and inputs entered are not collected or seen by the model. Session cookies persist so users typically need to sign in only once per site. The feature covers tasks such as booking appointments, dealing with landlords, insurance companies, government agencies, restocking supplies, and submitting business paperwork, and extends the agent's reach to authenticated dashboards, SaaS tools, internal portals, and paywalled content.
User adoption figures published August 31 by Kyle Reidhead show Codex and ChatGPT Work, OpenAI's autonomous agent that authenticates with third-party sites via a remote browser, grew from roughly 6 million to 25 million users in about seven weeks, adding approximately 5 million users every 10 days since the July 9 launch.
Reidhead also noted that coding agent users drive substantially more inference demand than standard chat users.
Simon Willison published a technical breakdown on August 30 arguing that ChatGPT Work, OpenAI's autonomous agent that authenticates with third-party sites via a remote browser without the model seeing credentials, is effectively two products: a cloud variant and a local desktop version.
The cloud version includes code execution with unrestricted internet access, a headless Chrome browser, a persistent shared filesystem, Cloudflare Workers deployment, sub-agent orchestration, and scheduled automations. Willison flags that the combination of private data access, exposure to untrusted content, and exfiltration channels forms what he calls the 'lethal trifecta' for prompt injection risk, and notes OpenAI has not publicly explained its mitigations.
OpenAI launched ChatGPT Work on August 26, an autonomous agent that operates inside login-gated sites by pausing so the user signs in through an isolated remote browser; session cookies then persist for repeat tasks and the model never handles credentials, with screenshots disabled during sign-in.
Targeted tasks include booking appointments, managing insurance, and submitting business paperwork; usage is metered on the same structure as Codex, and Enterprise and Edu admins receive workspace spend controls in the Admin Console.
OpenAI introduced Admin plugin for ChatGPT Work and Codex
Stampli case study published showing 3.16x speed improvement on Deep Finance product launch using Codex and ChatGPT Work
I send one email each morning with the stories that moved. If you would rather just read here, that works too.
Subscribe free