Pretty scary stuff.
The AI technology behind this must need a reasonable amount of data - voice recordings of the person it's mimicking - to build a realistic immitation though. In that acticle the mother suggests that it perfectly mimmicked the 'way her daughter would cry', to do that I must assume there must be voice recordings online of her daugher crying on youtube, tictoc etc.
I think this just highlights how important it is to control the amount of information you share online, especially those at risk groups. That's always been the case of course with things like address, age, bank details etc, but now we need to be careful with other types of info too, such as our/our children's voice data!