ChatGPT Work Can Now Log Into Password-Protected Websites And Complete Tasks For Users

A login page is no longer always a hard stop for ChatGPT Work.

OpenAI has introduced a cloud browser feature that allows the AI agent to access supported password-protected websites after the user signs in, then continue completing tasks within the authenticated session.

ChatGPT Work Can Now Handle Signed-In Websites

The feature gives ChatGPT Work access to a wider range of online tasks, including websites that previously required users to take over and log in themselves.

The cloud browser can read webpages, click buttons, fill in forms and complete supported tasks on both public and signed-in websites.

Users can ask the agent to handle jobs such as checking insurance information, rescheduling deliveries, cancelling flights, arranging appointments, managing utilities or working with online business systems.

The feature is available on web and mobile for ChatGPT Plus, Pro and Business users, and was included in OpenAI’s 25 August 2026 release notes.

Your Password Is Entered Outside The AI Model

ChatGPT does not ask users to type their login details into the conversation.

When the agent reaches a website that requires authentication, it pauses and opens a secure login form inside its remote browser.

Users can enter their username, password and two-factor authentication code themselves, or use a supported third-party password manager to autofill the details.

OpenAI says the credentials entered through this form are not visible to the model and are not stored by ChatGPT.

OpenAI also says sign-in requests are reviewed by an additional model for signs of phishing or deception before the agent continues.

How Does ChatGPT Stay Signed In?

After the user completes the login process, the cloud browser can retain the authenticated session through cookies.

This means ChatGPT Work may not need the user to log in again when it returns to the same website for another task.

The session operates separately from the user’s normal browser.

The cloud browser runs on a remote computer, so it does not automatically gain access to open tabs, saved passwords, cookies, extensions or browsing history stored on the user’s device.

Users can clear cloud browser data and manage website permissions when needed, including clearing sessions for individual sites through Settings > Cloud browser.

The Convenience Comes With A New Security Risk

The ability to remain signed in removes one of the biggest obstacles facing AI agents, but it also gives an agent continuing access to an account after the password has been entered.

OpenAI describes the experience as allowing users to “hand off a task and step away while it keeps working”.

That convenience creates a different security concern from password exposure.

While the AI may not see the password itself, it can operate within the authenticated session and potentially access information or perform actions available to the logged-in account.

This is particularly important for accounts containing financial information, private documents, emails, account settings or other sensitive data.

Users Still Control High-Stakes Actions

OpenAI says ChatGPT is designed to request confirmation before actions that could create financial, legal, account-related or other real-world commitments.

Some websites may also block automated browsers, while certain tasks can still require the user to take over and complete the final stage.

Users are advised not to enter passwords, security codes or payment information directly into the ChatGPT conversation.

Instead, those details should be entered through the secure sign-in process or takeover mode.

ChatGPT Work Is Moving Further Into Real Online Tasks

The update pushes ChatGPT Work beyond simply providing instructions and towards completing longer workflows on behalf of users.

Its ability to access signed-in websites means the agent can now continue working inside accounts after the user has authenticated, potentially removing repeated login steps from everyday online tasks.

The main trade-off is straightforward: users gain convenience by allowing ChatGPT to remain signed in, but they also need to consider how much control they are comfortable giving an AI agent over an active account session.