Nearly 200 Startups Ask Trump to Reject a Chinese Open-Weight AI Ban

Khanh Nguyen
Khanh Nguyen
(Updated: )
Listen to this article0 / 0
President Donald Trump during a White House Cabinet meeting, May 2026. (Photo: Win McNamee/Getty Images)

A newly formed startup trade group told the White House this week that cutting off Chinese open-weight AI models would kill American companies, not Chinese ones.

Kimi K3's Debut Set Off a Week of Escalating Signals From Washington

Moonshot AI's release of Kimi K3 on July 16 set off the current round of scrutiny. The model's strong showing on a coding benchmark was followed within days by reports that the administration was weighing new restrictions on Chinese open-weight models. By July 21, Treasury Secretary Scott Bessent said the administration would look into whether Chinese AI firms had improperly distilled American models. The next day, OSTP Director Michael Kratsios went further, alleging Moonshot AI used Anthropic's Fable model during Kimi K3's development through an internal distillation system, and that the company obtained Nvidia GB300 servers despite export controls. Moonshot AI has not publicly responded to either claim. Kratsios said the administration still supports open-weight development broadly, drawing a line between ordinary distillation and large-scale intellectual-property theft.

One Week, Five Moves: The Chinese Open-Weight AI StandoffTimeline of events from Kimi K3's launch through the Little Tech Association's letter and the White House's response, July 16 to July 22, 2026.One Week, Five MovesFrom Kimi K3's launch to the White House's non-answerJul 16Kimi K3 launches,tops a code benchmarkJul 21Bessent flags adistillation probeJul 22Kratsios allegesFable distillation, GB300sJul 22~200 startups sendthe Little Tech letterJul 22White House: no policychange announcedSource: Politico, Little Tech Association letter (Jul 22, 2026)

The Little Tech Association's Letter Asked for Safeguards, Not a Ban

The group behind Wednesday's letter, the Little Tech Association, is new — this appears to be its first coordinated push on a major AI policy fight. Its letter, addressed to President Trump, Commerce Secretary Howard Lutnick, and Kratsios, argued that American leadership depends on both strong U.S. open-weight models and continued access to open models already available worldwide, and it asked for targeted safeguards rather than a blanket restriction. Particle founder Suhail Doshi, a member of the group, warned that "there'll be hundreds of companies that instantly die" if access were cut off, adding that closed-model vendors like Anthropic would be the beneficiaries. Little Tech Association executive director Harry Godfrey framed the group's ask as finding the lightest-touch option that addresses security concerns without raising costs or limiting access. According to people familiar with internal discussions, senior White House and Cabinet officials debated the China AI-model question earlier in the week, and a blanket prohibition was reportedly not seriously considered.

The Letter, by the NumbersThree reference figures from the Little Tech Association's July 22 letter and the state of formal enforcement action.The Letter, by the NumbersAs of Politico's July 22 reportStartups who signed~200Backed by YC, ProtonOfficials addressed3Trump, Lutnick, KratsiosEntity List filings drafted0Per a person familiar, as of Jul 22Source: Little Tech Association letter; Politico, Jul 22, 2026

Commerce Has Drafted No Entity List Action, Leaving Two Paths Open

What happens next runs through the Commerce Department, and it has not moved. As of Wednesday, Commerce had drafted no plan to add Chinese AI companies to its Entity List, according to a person familiar with the matter. White House spokesperson Liz Huston said the administration has not announced any policy change and is focused on widening the U.S. lead in AI, while a White House official dismissed reports of pending action as speculation. That leaves the dispute split along a predictable line: larger, closed-model U.S. developers have pushed for tighter restrictions on Chinese AI developers on national-security and IP grounds, a stance that lines up with the separate distillation and export-evasion allegations Kratsios raised against Moonshot AI — allegations distinct from, and unresolved ahead of, any decision on a blanket access ban. Startups argue restricting models already downloadable worldwide would not slow China's progress but would raise their own costs, a dynamic that echoes how the end of subsidized AI pricing is already reshaping enterprise budgets. Whichever path the administration takes will also land differently depending on whether Beijing-based vendors keep pace on frontier capability, a question raised by one Chinese security firm's own claims of a Mythos-class model built after the export ban.

Two Unresolved Paths for Chinese Open-Weight AI PolicyA decision diagram showing the blanket-ban path startups oppose against the targeted-safeguards path they are requesting, both stemming from the same open-weight model question.Two Unresolved PathsNeither path has been formally chosen as of publicationChinese open-weight modelsBlanket banReportedly not seriously discussedTargeted safeguardsWhat the letter requestsStartups say hundredsof companies foldIP claims addressedwithout cutting access

Nothing here is settled. Moonshot AI has not responded to Kratsios's distillation and export-evasion claims, Commerce has not filed anything, and the White House has confirmed only that no decision has been made. What changed this week is that startups, not just frontier labs, are now on the record with Washington — and they are on the opposite side of the table from companies like Anthropic.

Comments (0)

Sort by:

No comments yet.

Be the first to share your perspective on this topic.