Continuous Approval

Continuous Approval is a setting that lets AI carry out an action right away without asking the user for confirmation each time, because permission was already granted in advance.

Think of it like handing documents to an assistant: instead of getting your signature every single time, you tell them upfront, "From now on, go ahead and handle this yourself without checking with me." Normally the assistant would draft something, show it to you for approval, and only send it after you sign off. With Continuous Approval turned on, that final review step disappears, and the assistant judges for itself and follows through to completion.

This saves the hassle of checking in every time, but it also means taking on the risk that something unwanted could slip through and go out by mistake. That's why this kind of feature is usually not the default — it's an option users have to turn on separately when they need it.

In the article, this appears in the description of ChatGPT's Apple Messages plugin. OpenAI's documentation states that "turning on Continuous Approval means ChatGPT loses the chance to do a final review before sending a message on the user's behalf." The easily misunderstood part is that this is not the default behavior. The default is to confirm before sending each time; Continuous Approval (automatic sending) is an option users must turn on separately.

When you come across a setting for automation or auto-execution in an AI assistant or agent feature, ask yourself first: "If I turn this on, will I still be asked to confirm each time, or will it run automatically without confirmation?" If the answer is the latter, that's a Continuous Approval-type setting. For hard-to-undo actions like sending messages or making payments, it's safer to think twice before turning this option on.