Maybe yes? Lets do s/AI/ML/g and we now that ML/DL is just stats. It's great at picking up patterns that are repetitive. The key is that it really generalizes well over things that are in fact repetitive and settle into a single function->domain response. Unfortunately humans are extreme creatures of habit. That habit creates patterns and therefore I suspect we have no home long term of privacy. Time to be open!
you don't need AI for that and,and it's pretty much already dead. Have you looked intk xkeyscore and the prism program? https://en.wikipedia.org/wiki/PRISM_(surveillance_program)
> Documents indicate that PRISM is "the number one source of raw intelligence used for NSA analytic reports", and it accounts for 91% of the NSA's internet traffic acquired
That coupled with malicious Tor relays(https://www.theguardian.com/world/2013/oct/04/nsa-gchq-attac...) makes practical low latency anonymity extremely difficult to achieve. You can have some confidentiality though.
Maybe you should/could ask if ML would enable "good enough" anonymity through intelligent route selection and avoidance and traffic fingerprint normalization.
ML is a double edged sword.