Check out: When China Sees All, Chinese AI Is Creating an Axis of Autocracy, by Ross Andersen (The Atlantic, Sept 2020 print edition) And Sheena Greitens analysis of it.
This article made me think that the US Defense agencies and freedom fighter hackers could be working on cheap devices to detect and disable surveillance cameras from afar(*), and on "prion" computer viruses(**). Such liberating technology could be put into the hands of oppressed people. I think China's use of AI for nefarious purposes would not be super difficult to foil because it relies on very fragile, networked devices.
(*) One could, for instance, tap into the surveillance camera network and inject a virus through there.
(**) By prion computer viruses I mean computer viruses specifically designed to surreptitiously degrade the behavior of artificial neural networks (ANNs) or other AI software. ANNs are very finicky and fragile devices that take a long time to train; certain subtle anomalies in that training will cause the ANN's efficiency to plummet, resulting in the loss of days or weeks of training. I've heard of software designed to mitigate biases and blind spots in ANNs by subtly altering the final weights of the ANN. Couldn't similar ideas be used to introduce biases and blind spots into ANN software? If done right, the virus could corrupt an ANN's training or final weights and disappear once its job is done, unbeknownst to the ANN owner.