США выбирают точечные, а не тотальные запреты китайских open-weight моделей

The Trump administration is leaning toward targeted bans on individual Chinese open-weight AI models instead of a blanket ban — that's according to a July 2026 report by The Decoder, citing national security considerations.

What exactly the US is planning

The US is considering banning specific Chinese open-weight models rather than all open-weight systems at once. According to The Decoder, the Trump administration views a selective approach as less damaging to the market and less exposed to accusations of blanket protectionism than an across-the-board barrier.

• The format of the restrictions is targeted, model by model, rather than a blanket ban on the entire open-weight class
• The stated reason is national security considerations
• OpenAI and Google DeepMind signed an open letter against regulating open-weight models
• OpenAI and Anthropic, according to the source, are privately lobbying for the same restrictions
• The status is reported as a plan (reportedly), not as an adopted decision

Why OpenAI's position looks contradictory

OpenAI is simultaneously arguing both against and for regulating open-weight models. Publicly, the company signed an open letter together with Google DeepMind against such restrictions — a reaction to public pressure. But privately, as The Decoder reports, OpenAI and Anthropic continue to lobby for the introduction of those very same bans.

According to

The Decoder, the Trump administration is leaning toward banning specific Chinese models rather than all open-weight systems at once, citing security risks.

This gap between the public and behind-the-scenes positions is explained by a dual motivation: on one hand, security risks from foreign models, and on the other, the direct business interests of American labs, which stand to benefit from restricting strong Chinese competitors like DeepSeek and Qwen.

What's behind the decision

Two factors lie behind the choice of a targeted approach: security and competition. Open-weight models can be downloaded and run locally without any control by the developer, so regulators are concerned that American companies and government agencies will make widespread use of Chinese open source.

At the same time, restricting Chinese models plays into the hands of OpenAI, Anthropic, and Google, reducing pressure from cheap and powerful open competitors. It is precisely this overlap between security arguments and commercial benefit that makes the lobbying in July 2026 particularly notable.

What it means

Washington is feeling out a middle option between the freedom of the open-weight ecosystem and an outright ban: striking in a targeted way at specific Chinese models rather than the entire class. For the industry, this is a signal that regulating open models is ceasing to be a theoretical debate and is turning into a tool of the US-China technology standoff.

Frequently asked questions

Is the US planning to completely ban Chinese AI models?

No. According to The Decoder, the Trump administration is leaning toward targeted bans on individual Chinese open-weight models rather than a blanket ban on the entire open-weight class.

Why does OpenAI's position look contradictory?

OpenAI, together with Google DeepMind, publicly signed an open letter against regulating open-weight models, but, according to the source, is privately continuing to lobby for the same restrictions together with Anthropic.