Trump won’t put the brakes on AI because he wants to beat China. How big a threat is Beijing?

The US approach to AI is a zero-sum game: win or lose against China. Trump's stance directly opposes calls from American AI leaders like Anthropic and OpenAI for development slowdowns. Washington sees AI as a strategic asset to protect and export to allies. This contrasts sharply with Beijing's push for open collaboration, a move that likely reflects its current lag in advanced AI capabilities.

Chinese AI labs are closing the gap with US competitors. DeepSeek shocked the industry last year with a model nearly matching leading American systems using fewer computing resources. Since then, Alibaba, Z.AI, Minimax, and Moonshot have unveiled similarly capable models. This rapid iteration challenges the dominance of US frontier companies, but it also fuels US accusations of intellectual property theft, specifically distillation from American models.

The US alleges "industrial-scale" theft by Chinese AI firms. Anthropic's report details attempts by alleged Chinese government-affiliated actors to misuse its systems for surveillance. One example cites Moonshot's Kimi model rerouting user queries to Claude, exposing an attempt to divert requests. Beijing rejects these claims, but the accusations will escalate tensions at the Trump-Xi summit in two weeks. This will pressure Chinese AI companies to demonstrate independent innovation.