Exporting the Surveillance State via Trade in AI is an interesting paper from a group of economists from MIT, Harvard, and Oxord studying "the global diffusion of surveillance AI technology [meaning facial recognition], and in particular, the role played by China''. The results they find are worth reviewing although not very surprising: China is more likely to export surveillance AI than other countries, autocracies are more likely to import it from China, and imports are more likely during political unrest.

I saw this paper through NBER, so it's worth noting that this is a international trade economics study — I think their empirical findings are plausible and entirely in agreement with China's own geopolitical roles, but it's more "relative advantage vs the US" than "monopoly," and the map is incomplete without corporate actors in other countries — many of them First World democracies — providing technology and services to some in the same market, never mind internal customers of dubious bona fide. And that's ignoring military-only technologies, of course. (I'm focusing only on aspect of the paper's conclusions; its analysis and data go far beyond this, and it's worth reading and digesting on its entirety.)

The remarkable thing is how... traditional all of this is. Countries with an edge on technologies of political surveillance and control have always sold them to those without, both as business and as diplomacy. Current AI surveillance technologies aren't even particularly good, although they carry a sci-fi aura that's not without its value.

A good mental exercise: If you wanted to "export" generally prosperous democracy through trade in AI, would you know what to sell? Would you know what to build? What would it look like?

None of those are easy questions. The world is still figuring them out.