A Register opinion piece argues that Moonshot AI's new Kimi K3 model shows Chinese open-weight AI has become genuinely competitive with US frontier systems, even as American firms increasingly frame Chinese models as national security threats rather than engage with their technical merits. The author notes this pattern recurs every few months, following DeepSeek R1's market-shaking debut last year, and suggests the "Chinese AI" framing conveniently benefits companies like Anthropic and OpenAI, which can use fears of foreign models to bolster arguments against open-weight AI generally while facing less scrutiny of their own practices.

At 2.8 trillion parameters, Kimi K3 is described as the largest open-weights model built to date, reinforcing perceptions of it as a genuine frontier system, though its scale puts it beyond most enterprises' practical reach. The piece contrasts this with recent US developments: GPT-5.6's release was reportedly delayed by government concern, and Claude Fable 5 was briefly pulled offline over security worries that turned out to be relatively minor. It also notes Anthropic chief Dario Amodei's past accusations that Alibaba used Claude to improve its Qwen models via distillation, juxtaposed against Anthropic's own $1.5 billion settlement over allegations of training Claude on pirated books. The White House is reportedly facing calls to restrict access to Chinese models but has not taken formal action.