Back to feed
News Story
THE DECODER
1 sources

US reportedly favors selective bans over blanket restrictions on Chinese open weight models citing security concerns

The Trump administration is reportedly planning targeted bans on Chinese AI models instead of a blanket restriction, citing security concerns. This comes after OpenAI and Google DeepMind publicly opposed regulating open-weight models, while privately lobbying for restrictions. The article highlights the tension between public stances and private business interests in AI policy.

SynthePulse Insight · AI deep reading

U.S. AI Model Controls on China: The Game of Selective Bans and Uncertainty

Version 1 · 1 source

The Trump administration is reportedly leaning toward targeted bans on specific Chinese open-source AI models rather than a blanket crackdown. This policy choice reflects both national security considerations and the complex stance of U.S. tech giants between public statements and private lobbying.

  • The Trump administration plans targeted bans on specific Chinese AI models, not a blanket ban.
  • OpenAI and Google DeepMind signed an open letter opposing regulation of open-source models, but OpenAI and Anthropic are privately lobbying to restrict Chinese models.
  • Current Chinese models (e.g., Kimi K3) still significantly lag behind leading Western models on cybersecurity benchmarks.
  • Many companies supporting the open letter have commercial interests in Chinese open-source models while also seeking to curb the market power of Anthropic and OpenAI.
Open section navigationPolicy Direction: Selective Bans Rather Than Blanket Crackdown

Policy Direction: Selective Bans Rather Than Blanket Crackdown

According to a New York Times report, the Trump administration is leaning toward targeted bans on specific Chinese AI models rather than a blanket ban, citing that national security concerns may be confirmed over time. This policy direction indicates that the U.S. government, while restricting Chinese AI technology, is trying to avoid overly damaging its own commercial ecosystem that relies on open-source models.

The report notes that even recently released models (e.g., Kimi K3) still significantly lag behind leading Western models on cybersecurity benchmarks. This fact provides a technical basis for selective bans—not all Chinese models pose equal threats.

Tech Giants' Public Stance and Private Actions

OpenAI and Google DeepMind signed an open letter opposing regulation of open-source AI models. Many large tech companies supported the letter, most of which have commercial interests in Chinese open-source models—Microsoft and Google sell access to these models through their services. Google DeepMind itself also develops open-source models, including the successful Gemma series.

However, the New York Times reports that Anthropic and OpenAI are privately lobbying to restrict Chinese open-source models, despite OpenAI signing the open letter. These two companies face price pressure from cheaper Chinese models. This dual stance of publicly opposing regulation while privately pushing for restrictions reveals deep contradictions between commercial interests and security concerns.

Intertwined Interests in the Open-Source Model Ecosystem

Many supporters of the open letter have direct commercial interests in Chinese open-source models. Microsoft and Google sell these models through cloud services, while Google DeepMind's Gemma series itself is a successful open-source model. These companies also seek to curb the market power of Anthropic and OpenAI.

This complex interest landscape means that any restrictions targeting Chinese models could affect U.S. companies' businesses. Selective bans may be a compromise: addressing security concerns without completely severing commercial interest chains.

Credibility boundary

The core information in this article originates from a New York Times report, as relayed by THE DECODER. THE DECODER is a secondary source; its relay of the New York Times report should be considered as claims from the original report, not independently verified facts. Details about Kimi K3's benchmark performance, private lobbying by OpenAI and Anthropic, etc., all come from that report and have not been independently verified by third parties.

Insight takeaway

U.S. AI model controls on China are moving toward selective bans, driven by both national security and commercial interests. The split stance of tech giants between public letters and private lobbying highlights the complexity of regulatory games. The current technical gap of Chinese models provides a window for selective bans, but as model capabilities improve, policies may tighten further.

Sources for this version

  1. US reportedly favors selective bans over blanket restrictions on Chinese open weight models citing security concernsTHE DECODER

Primary report

THE DECODER

Primary source