It's probably easier/faster to just have AI generate vector art then use a tool to automatically vectorize it.

Have a look at https://clipasso.github.io/clipasso/

Diffusion models can be trained to learn distributions over more or less arbitrary spaces of signals.