The “dots” associated with ChatGPT refer to those intelligent systems that continue to perform their tasks in the cloud-based browser even after you log out. Now, they have gained an additional capability: they can read one-time verification codes sent to your email account and complete the login process on their own.
When a verification code is required, you will be asked whether it is allowed only this once or whether it will be automatically filled in in future instances. In the official demonstration, a user used their email address to log into the restaurant reservation app Resy.
Prior to this, OpenAI’s help pages always required users to complete such verifications on their own, or to manually perform the necessary operations using a web browser.
James Sun, who is in charge of the chatbot capabilities of ChatGPT as well as the browsing features of Codex at OpenAI, announced this change. From a product development perspective, this step was inevitable: if intelligent systems were truly intended to assist people in performing tasks, logging in became an essential requirement.
It’s truly convenient – now, tasks like placing orders, changing reservations, and checking order details can all be completed in one go, without having to wait for anything in the middle.
The very purpose of a verification code is to prove that “it is you who is performing the operation.” If you hand this function over to an intelligent system, you are essentially relinquishing this layer of protection. Once the verification code is manipulated through means such as the insertion of misleading prompts, that system will then have access to the login credentials for your multiple accounts.
The suggestion is: choose “Only this time” instead of the more convenient option “Always allow”.