Trump Under Pressure: Startups Worried About Access to Chinese AIs

American Startups Concerned About Restrictions on Chinese AIs

Nearly 200 American startups have recently expressed their concern over the possibility that the Trump administration may restrict access to artificial intelligence (AI) models developed by Chinese companies. These companies, which include influential players from Silicon Valley like Proton and Y Combinator, believe that such a measure could severely harm innovation and competitiveness among American startups.

The Little Tech Association, a newly formed collective, has taken the initiative to send letters to the Trump administration advocating for the maintenance of this access. These letters, sent on Wednesday, warn against the consequences of cutting off access to open-weight AI models, which are models whose parameters are publicly available. Chinese companies such as Moonshot AI and Alibaba are behind these models.

A Call for Targeted Protective Measures

In a letter addressed to Michael Kratsios, director of the Office of Science and Technology Policy, startup founders emphasized the importance of the United States maintaining global leadership in AI. They stressed the need for top-tier American models while keeping access to open models available internationally. Rather than imposing blanket bans, they advocate for more targeted protective measures.

The founders warn that restricting access to Chinese models would only harm American startups without preventing the spread of these technologies. Suhail Doshi, founder of the AI infrastructure startup Particle, stated in an interview that this could lead to the disappearance of hundreds of companies, which would then be forced to turn to more expensive solutions offered by giants like Anthropic.

The Urgency of the Situation

The issue has taken on new dimensions after reports suggested that the Trump administration was seriously considering banning Chinese AI models. This possibility was raised shortly after Moonshot AI launched its Kimi K3 model, which is considered very powerful.

Liz Huston, a White House spokesperson, asserted that the United States intends to remain at the forefront of AI innovation and that President Trump is determined to maintain this advantage. She added that the administration is ready to intensify its efforts to widen the gap between America and the rest of the world in this field.

Behind-the-Scenes Discussions

A White House official, who wished to remain anonymous, clarified that any policy announcement would be made directly by the administration, labeling current reports as unfounded speculation. Nevertheless, these rumors have caused some panic among Silicon Valley startups, which rely on open-weight Chinese models to develop their products at lower costs.

Discussions took place on Monday evening between senior members of the White House and the Cabinet, although the idea of a blanket ban was not seriously considered. However, the administration has ramped up its rhetoric against Chinese AI developers, with accusations of intellectual property theft.

Accusations of Model Distillation

Scott Bessent, Secretary of the Treasury, stated on Fox Business that the administration would investigate the possibility that Chinese companies have illegally used American models to develop their own technologies. He mentioned the possibility of sanctions against companies involved in such practices.

Michael Kratsios claimed that Moonshot AI had distilled Anthropic's Fable model to create K3, using a sophisticated platform to evade detection. He also noted that Moonshot had acquired Nvidia GB300 servers, despite a sales ban to Chinese entities.

A Debate on Security and Innovation

Kratsios reiterated the administration's support for a free and fair development of AI while distinguishing legitimate model distillation from large-scale theft. Moonshot has not yet publicly responded to these accusations.

To date, the Department of Commerce has not included Chinese AI companies on its list of entities subject to export restrictions. The Little Tech Association's stance against banning open-weight models reflects a growing divide within the AI industry.

Harry Godfrey, executive director of the Little Tech Association, stated that policymakers should take a measured approach to avoid hindering American innovation while addressing security concerns. He suggested that the solution lies in precise measures that do not harm access or competitiveness for startups.