U.S. Targets Chinese AI Models With Selective Ban

The Trump administration is weighing a narrower approach to Chinese open-weight AI models than many expected, according to the New York Times. Rather than a blanket ban, officials are reportedly leaning toward targeted restrictions on specific models, a move framed around national security concerns tied to cybersecurity risks. Models like Moonshot AI's Kimi K3 still lag well behind leading Western systems on cybersecurity benchmarks, but that gap is exactly what worries some in Washington looking ahead.

The timing matters. Just as the White House moves toward tighter controls, OpenAI and Google DeepMind put their names on an open letter opposing regulation of open-weight models generally. That's a notable position for two companies with plenty to gain from the status quo, and it sits awkwardly next to reports that OpenAI and Anthropic have been lobbying behind closed doors for restrictions on Chinese open models specifically.

Both companies are also under real pricing pressure from cheaper competitors coming out of China. The letter has drawn broad backing across the industry, revealing a tangle of competing motives among the companies pushing it, some aimed at Beijing, others at Anthropic and OpenAI's grip on the market.