As Customer said, the detection zone is 100% local on the camera to determine when something is a qualified “event” from motion there; and PD AI processing is done 100% in the cloud on the ENTIRE video (it doesn’t know about detection zones).
It means you had some motion inside your detection zone (maybe a leaf blowing, or plant moving, or a light), so it uploaded an event to the cloud to analyze, and the AI detected a person somewhere on the video.
I have other companies’ cameras and most work quite similarly (any motion in the zone triggers analysis of the entire picture), so this is fairly normal.
If you want the AI to ignore certain areas too, you can try to physically block them from being recorded such as putting something partially over the lens on that side, or put up cardboard on the side, similar to the blinders they use for horses to block their view. Or rearrange your cam to not cover those areas you don’t want in view.
I have otherwise found that the detection zone works quite well