Sluggish adoption can’t be blamed on overregulation, and restricting access to US frontier models is no defence against open-source Chinese rivals
3-MIN READ3-MIN ListenBruno S. SergiandKevin K.F. ChenPublished: 8:30pm, 17 Aug 2026Last month, two Chinese companies released artificial intelligence models most people outside the industry know little about: Kimi K3 by Moonshot AI and Qwen3.8-Max by Alibaba Group Holding (which owns the South China Morning Post). The markets reacted even before publicly auditable evidence about the performance claims – detailed benchmark table, model card and licence – was made fully available.In the weeks that followed, global chip stocks shed about US$3 trillion in market value.
That reaction, rather than the models, is the story. Obviously, nobody was repricing Chinese engineering – the benchmarks were not yet verified. What was being repriced was scarcity: the long-held assumption that the most capable systems would remain expensive, American and controllable. Portfolios had been built on that assumption. It is worth less than it was.
The restrictionists believe American advantage can be preserved by denial. The accelerationists believe it is being smothered by regulation. Both camps are arguing about the wrong obstacle.
There is very little regulation to blame. The United States has no comprehensive federal AI statute, despite the recent push to create a self-regulatory body in the style of the Financial Industry Regulatory Authority for frontier and AI models. Almost nothing is stopping an American hospital, insurer or car plant from deploying these systems.
