OpenAI has expanded ChatGPT Work's cloud browser to support signing in to websites that require authentication. The capability lets a Work task pause at a login screen, request the user's input, and continue after authentication across web and mobile tasks. According to OpenAI's cloud browser documentation, credentials entered during this process go directly to the remote cloud browser. They are not visible to the ChatGPT model and are not stored by ChatGPT.
The update matters because many useful business workflows happen behind a login. Teams often need information from dashboards, account portals, web applications, and other signed-in services. ChatGPT Work can now potentially handle parts of those workflows in its cloud browser while keeping the actual credential entry outside the model's context.
How authenticated cloud browsing works
When a ChatGPT Work task reaches a sign-in page, the service can prompt the user to authenticate through a secure sign-in form. OpenAI says an additional review model evaluates the situation before the sign-in request is presented. Users can also preview the sign-in page and the live form before continuing.
The key distinction is where the login occurs. The user enters credentials into the remote browser environment that is running the task, rather than supplying them to the model in the chat. That design means ChatGPT can resume its work after the user signs in without receiving the plaintext username or password.
Sessions persist, but remain separate from local browsing
After authentication, the session can persist for future ChatGPT Work tasks until it expires. This can remove the need to sign in again for every eligible task. The cloud browser uses its own cookies and data, separate from a user's local browser, so a local browsing session is not simply transferred into ChatGPT Work.
| Task state | Before authentication | After authentication |
|---|---|---|
| Website access | The task stops when it reaches a required sign-in page. | The task can resume in the cloud browser with access to the authenticated session. |
| Credential entry | No login credentials have been entered in the remote browser. | The user enters credentials directly into the secure remote-browser form, not into the model. |
| Future tasks | A signed-in session is not yet available. | The session can persist until it expires, avoiding repeated authentication for each task. |
This is a material extension of cloud browsing because it makes authenticated web work possible rather than limiting tasks to pages available without a login. It also fits the broader ChatGPT Work workflow described by OpenAI, including browser integration in the desktop app, connections to tools and plugins, and persistent project management across devices.
Availability and website limits
Access is limited to paid ChatGPT plans in supported regions, and availability can vary across workspaces. OpenAI's documentation also makes clear that a cloud browser is still subject to each website's policies. Some sites may block cloud-browser access, in which case the user may need to handle the login manually in their own browser.
Authentication can also involve steps that no browser agent can bypass on its own. For example, a site may require two-factor authentication. Users should expect to complete such checks and should review sensitive actions before allowing a task to proceed.
What the update means for business automation
For businesses, the practical opportunity is not automatic access to every account. It is the ability to reduce repetitive navigation and information-gathering work in workflows where a person can securely authenticate and retain control of consequential steps.
Potential uses supported by the new authenticated browsing model include:
- Pulling information from dashboards that require an account login.
- Completing routine account-bound browser tasks after a user has authenticated.
- Continuing multi-step web workflows without re-entering credentials for every Work task while the session remains active.
- Combining signed-in browser activity with connected tools, plugins, and ongoing project context where those capabilities are available.
The credential handling is especially important. Asking staff to paste passwords into an AI chat would create an obvious security concern. OpenAI's described flow is designed to avoid that specific exposure by sending credentials into the cloud browser rather than to the model. That does not remove the need for sensible access decisions. A team still needs to decide which accounts are appropriate for a browser task, who can authorize sign-ins, and which actions deserve human review.
The most suitable early workflows are likely to be repeatable, low-risk tasks with a clear outcome, such as retrieving account information or progressing through a familiar portal process. Workflows involving irreversible financial, legal, customer, or administrative actions require more caution. Website restrictions, expired sessions, and two-factor authentication can also interrupt an otherwise useful automation.
Authenticated browser tasks can reduce the manual handoffs that slow routine operations, but only when access and review steps fit the workflow. Scalevise helps teams assess where AI can safely automate dashboard work, account-bound processes, and tool connections without overcomplicating everyday operations. Talk to Scalevise about a practical AI automation project to map a useful first workflow, clarify the human checks it needs, and move from experimentation to a workable process.
Frequently Asked Questions
What is new in ChatGPT Work's cloud browser?
ChatGPT Work can now support signing in to websites that require authentication. A task can pause at a login page, let the user authenticate in a secure remote-browser form, and then resume where it left off.
Can ChatGPT see or store my website password?
OpenAI says credentials entered in the sign-in flow are sent directly to the remote cloud browser. They are not visible to the ChatGPT model and are not stored by ChatGPT.
Will users need to sign in for every ChatGPT Work task?
Not necessarily. Once authenticated, the session can persist for future Work tasks until it expires, so users may not need to re-authenticate for every task.
Can ChatGPT Work sign in to any website?
No. Website access remains subject to each site's policies, and some sites may block cloud-browser access. Two-factor authentication may also require user participation, and manual login in a local browser may sometimes be necessary.
Who can use authenticated cloud browsing in ChatGPT Work?
OpenAI says the capability is available on paid ChatGPT plans in supported regions, with rollout differences possible across workspaces.
Conclusion
ChatGPT Work's authenticated cloud browsing expands the kinds of web tasks the service can support by allowing users to sign in without exposing credentials to the model. Persistent sessions can make repeat tasks more practical, but website policies, authentication checks, and human review remain important limits. For teams, the value will depend on selecting routine signed-in workflows where the time saved outweighs the setup and oversight required.
