ChatGPT can now do things on your behalf without seeing your login details

ChatGPT can now check your insurance, cancel flights, or reschedule deliveries for you, all seemingly without seeing your login details.

Announced by OpenAI on X, the AI assistant will be able to complete tasks for users that they’d usually have to open a new browser and log in with a username and password to do. (And please don’t tell us you’ve been dropping your login details into AI chatbot conversations. Stop that now.)

This Tweet is currently unavailable. It might be loading or has been removed.

Available on mobile and web for Plus, Pro, and Business users, the update functions within the GPT‑5.6-powered ChatGPT Work, a task management agent within the chatbot that pulls workflows into one spot.

However, the tool is focused beyond boardrooms. According to OpenAI, ChatGPT Work can now do things like set up utilities for your home, book a passport or medical appointment, buy or cancel plane tickets, or rearrange a package’s arrival, all without viewing your login details. This moves the chatbot beyond its current in-chat booking abilities into deeper, more serious purchasing waters.

But how is ChatGPT actually blocked from seeing your username and password?

On OpenAI’s help page, updated with the announcement, the company says ChatGPT Work functions through a cloud browser that can “read web pages, click buttons, enter information into forms, and carry out steps on supported public and signed-in websites.” This means you’re not putting your details directly into the Large Language Model (LLM) powering the chatbot, and again, it’s unwise to drop these into your chats.

You’ll need to start a task within ChatGPT, write your request as a prompt, and the chatbot will pause the conversation to generate a “secure form” for logging into a website through a remote browser. You might need to approve the website to ChatGPT and complete any two-factor authentication. According to OpenAI, “the username and password entered [in the secure form] are not visible to the model, and ChatGPT does not store those sign-in credentials.”

The company adds that all sign-in requests are reviewed by an additional model for “signs of phishing or deception.”

You can also control which websites ChatGPT has access to in Settings > Cloud browser, where you can select options to “Always ask,” “Auto approve,” or “Always allow.” And some websites might restrict access from AI agents, so you might run into some that won’t work with this automation.

Once you’ve done all the approvals and you’re logged in, ChatGPT will unpause the conversation and move to complete requests. Notably, ChatGPT will ask you for confirmation in the chat “before actions that could be hard to reverse or create a financial, legal, account, or other real-world commitment, such as confirming a booking or making a payment.” So, if you’re about to drop cash on a flight, you’ll have to give it the green light.

ChatGPT users can already connect their bank accounts and medical records to the platform, so this feels like just another step of handing the keys over to AI.

Disclosure: Ziff Davis, Mashable’s parent company, in April 2025 filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.