The images that are flagged by such scanning, local or server side, will have to be manually verified to avoid false persecution. Someone will have to look at the private images you’ve sent that might get flagged.
These systems have huge margins of error and are incredibly inaccurate, so there will be a significant task in manually verifying everything. And do you trust some government random employee (or just the departments general IT practices or ability to not be hacked) with not leaking your nudes or personal images? I sure as hell don’t.
And even if this is handled perfectly and all government employees are super super honorable standup citizens that never do anything slightly wrong ever…There are still malicious governments that persecute minorities, I doubt they will handle these backdoors in digital privacy very well.