It’s physically impossible to perform any type of analysis on a low resolution video and end up with a massively* upscaled and accurate result no matter what sort of technology you use. AI just guesses from what it can see and what it’s been trained on. It’s the digital equivalent of hiring an artist to create a higher quality rendition of a picture.
*There are some tricks that can be used to “enhance” images and especially videos, but not so much that it’s like having optical zoom. For example, a bunch of noisy video frames of the same thing can be combined to get a less noisy result.
It literally happened in the US with period tracker app data getting subpoenaed in a state with an abortion ban.