That's not what the UK is demanding. They want client side scanning malware that breaks DRM, circumvents encryption and VPNs, and bypasses other security features in order to scan everything visible on your screen.
It absolutely doesn't. However, the argument doesn't work when it's about connecting the "is the user a kid" bit to the existing and constantly running object recognition (phone cameras already run skin detection all the time to set white balance), so people invent "third parties" and "report people to authorities".
But "Is the user a kid" is already a switch that I (a parent) switch on in the device and that the kid in question can't switch off. That bit seems like a solved problem?
Why would anything else even be needed in that space? The interest of parents and tech companies likely align here.