Why don't email apps use AI to avoid sending to the wrong person?

Seems like most everyone has experienced this. We have the confirmation button, but that only helps a little, it eventually gets done without thinking.

AI could use some metrics to detect how likely it is that you have the correct sender. If you don't talk to them often, or if the list of topics discussed contains something new you haven't sent that person, it could display extra warnings.

It's simple enough that it could even be done purely with on-device local AI.

At the very least, the confirmation screen should always show a summary, with the profile pictures of all recipients, and thumbnails of any attachments.