This morning's lead was the White House accusing China's Moonshot of stealing Anthropic's model. This is the other half of the same argument — and it's coming from inside the American tech industry.

Nearly 200 startups have told the Trump administration not to cut U.S. developers off from Chinese open-weight AI models. The letters, sent Wednesday to President Trump, Commerce Secretary Howard Lutnick, and others, were organized by a newly formed group called the Little Tech Association — and by the accounts circulating, the signatories include names like Proton and Y Combinator. It is the first time the broader startup community has moved as a bloc on one of the administration's most closely watched AI questions.

What they're pushing back against

The question on the table in Washington is whether to restrict U.S. access to the increasingly capable open-weight models coming out of Chinese labs — Moonshot's Kimi, Alibaba's Qwen, DeepSeek. Two mechanisms are in play. The one reported as most viable in Congress is a federal procurement ban: government agencies would be prohibited from buying or running Chinese AI models. And it's already begun in the national-security corner — the Fiscal Year 2026 National Defense Authorization Act directs the Department of Defense and the Director of National Intelligence to remove and exclude DeepSeek's AI from their devices, systems, and contractors.

So this isn't a hypothetical. The hawkish direction is set; the fight is over how far it goes.

The founders' argument

Their case has two prongs, and both are worth taking seriously. "American leadership requires two things," the letter reads: "world-leading American open-weight models and continued access for U.S. builders to open models already available worldwide." They're asking for targeted safeguards rather than sweeping restrictions.

The sharper point is the one about enforceability. Open-weight models are, by definition, already downloaded — the weights are sitting on hard drives and mirrors all over the world, and you cannot un-release them. A ban on Chinese open-weight AI wouldn't make Kimi or DeepSeek disappear; it would mostly stop law-abiding American companies from using them, while doing little to the models' actual availability. In that reading, the restriction lands hardest on exactly the U.S. startups it's meant to protect.

Why it matters

Put the day's two stories side by side and you get the real shape of the debate. The government's instinct, sharpened by accusations that Chinese labs are distilling American models, is to wall off Chinese AI. The startup community's counter is that walls around downloadable software mostly trap the people inside them.

It's a genuine policy fault line, and unusually, it doesn't run neatly along China-versus-America lines. It runs between American security hawks and American builders — one side worried about dependence on Chinese technology, the other worried that the cure kneecaps the domestic industry it's supposed to defend. With a procurement ban already moving and the NDAA precedent set, this is not a debate that stays abstract for long.