The Trump administration is showing renewed signs it could restrict access to cutting-edge Chinese AI models, following the release of Moonshot AI’s Kimi K3, a 2.8 trillion-parameter open-weights model that matched or surpassed US frontier models on key benchmarks at a fraction of the cost.
According to Axios and multiple industry sources, parts of the administration previously attempted to impose de facto bans on foreign open-source models, but those efforts were shelved by officials wary of stifling innovation. The emergence of Kimi K3, which achieved an 81.2 score on the FrontierSWE coding benchmark, ahead of both OpenAI’s GPT-5.6 Sol and Anthropic’s Claude Fable 5, has reignited those efforts.
The Commerce Department previously considered adding several Chinese AI labs to its Entity List, a step that would cut off US access without a government license. The National Security Agency and the White House Office of the National Cyber Director also discussed issuing an advisory on the security risks of Chinese AI models, effectively discouraging US firms from using the technology. Separately, the White House explored an executive order requiring US companies to guarantee security and accept liability for breaches before hosting Chinese models.
All these measures were shelved amid internal disagreement. But the departure of Sriram Krishnan, a former White House senior policy advisor who had opposed such restrictions, has shifted the balance. National security hawks, citing growing cybersecurity concerns, are now pushing for action.
However, an outright ban faces practical obstacles. Kimi K3’s model weights are already publicly downloadable, making enforcement difficult. Many US developers and enterprises have adopted Chinese open-source models because they offer frontier-level performance at dramatically lower prices, DeepSeek V4 Pro, for instance, costs under US$1 per million tokens (approximately £0.80) compared with US frontier models that can charge US$50 for comparable outputs.
A softer approach may prove more durable. Sources familiar with government discussions described a strategy combining procurement rules, Entity List threats, and public pressure campaigns to highlight potential backdoors in Chinese models, rather than pursuing a formal ban. The goal is to push US firms toward domestic alternatives while encouraging a more competitive American open-source ecosystem.
White House AI adviser David Sacks weighed in on the debate on X, warning that “the leading closed labs, already a duopoly in terms of AI model revenue, want the government to eliminate their open-source competition.”
Sources: The secret Trump administration battle to fight Chinese AI (Axios, July 2026); Uh-oh, the Trump admin doesn’t like China’s Kimi K3 AI model (Neowin, July 2026)