The AI Blame Game Just Went Global

Businessman refusing an envelope at a desk — Photo: PreciousJ / Shutterstock

China’s new claim that U.S. companies quietly train their artificial intelligence on Chinese models turns an already tense tech feud into a two-way fight over who is really stealing from whom.

Story Snapshot

China’s commerce ministry says many American AI firms “distilled” Chinese models during research and training, without naming specific companies.
U.S. officials and companies have already accused Chinese labs of industrial-scale copying of American AI models using distillation attacks.
American startups are increasingly building products on cheap or free Chinese open models, deepening dependence on a strategic rival.
Both governments now frame AI model reuse as a national security issue, while ordinary Americans see another example of elites playing by their own rules.

China’s New Accusation: U.S. Firms Distilled Chinese AI Models

Beijing’s Ministry of Commerce announced that it “understood” many American artificial intelligence companies had distilled Chinese models during their research, development, and training work. Officials used careful language. They did not name specific firms, show logs, or publish technical proof. But they claimed this behavior was widespread and part of how U.S. labs built their own systems. The ministry said Washington’s threats to sanction Chinese AI companies “lack factual basis and legal grounds” and amount to “artificial intelligence hegemonism.”

The Chinese statement came just days after reports that the Trump administration is weighing bans or tight limits on foreign open source models, especially those built through distillation on American systems. Beijing clearly wanted to flip the script. By accusing U.S. companies of similar conduct, China is trying to paint Washington as a hypocrite that enjoys open global data and code, but punishes rivals who do the same. That message plays to long-standing anger, on left and right, that global rules often favor big players and powerful governments.

U.S. Accusations Against Chinese Labs Set the Stage

For months, American officials and executives have warned that Chinese labs are copying U.S. frontier models at “industrial scale.” A White House tech memo said foreign entities in China run large campaigns using many fake accounts and special tools to pull outputs from American systems and then distill them into cheaper local models. Distillation here means training a smaller “student” model on the answers of a bigger “teacher” model, so the student learns the same behavior with lower cost.

Companies like Anthropic and OpenAI have gone further and given numbers. Anthropic told U.S. senators that Chinese firms DeepSeek, Moonshot, and MiniMax used tens of thousands of accounts and millions of chats with its Claude model to extract capabilities. OpenAI told lawmakers that DeepSeek built code to bypass access limits and automate collection of outputs for distillation. These claims helped drive talk of sanctions, new laws, and tighter rules on foreign access to U.S. AI services. They also fed fears among Americans that their own government could use “national security” to pick winners and losers in tech.

American Dependence on Cheap Chinese AI Models

While both sides talk about theft, U.S. firms are quietly relying more on Chinese models because they are cheap and easy to use. Reporting shows a rising number of American startups now build products on open Chinese models that can be downloaded and modified for free. These models often match the power of mid‑tier U.S. systems but cost far less to run, which matters a lot to smaller companies that cannot afford pricey American platforms. Some U.S. businesses are even centering their tools on these foreign systems.

U.S. lawmakers are starting to worry about this trend. Congress is probing the growing use of Chinese models by American companies and looking at options to curb it. Ideas include rules that would discourage any business that wants government contracts from using Chinese AI, and new efforts to spread information about possible risks or hidden weaknesses in those models. This debate lands in a country where many citizens already feel that trade and tech policy have shipped jobs and power overseas, while ordinary families fight inflation and stagnant wages at home.

Double Standards, Distillation, and the Deep State Worry

Both Washington and Beijing now use the word “distillation,” but they frame it differently. American officials say the problem is “unauthorized, industrial‑scale distillation” that breaks terms of service and undermines safety. Chinese officials answer that sanctions threats “adopt double standards” and show U.S. “hegemonism” in AI. Distillation itself is a normal technical method. Many labs use it on their own models. The real fight is about when reuse becomes theft, and who gets to decide.

The US-China tech war has moved from chips to code.

China is threatening "all necessary measures" if the US sanctions its AI firms. The flashpoint: Kimi K3, a 2.8T-parameter model allegedly distilled from Anthropic's tech. DeepSeek and MiniMax are also under scrutiny.

— orbitant (@orbitant) July 27, 2026

For Americans across the political spectrum, this back‑and‑forth can feel familiar and exhausting. Big tech firms accuse foreign rivals of cheating, while foreign governments accuse U.S. giants of abuse. Both sides demand new powers for regulators and national security agencies. Yet there is still no clear public evidence showing exactly which U.S. companies, if any, misused Chinese models, or a simple rule book that ordinary people can read and trust. That gap feeds the sense that the “deep state” and global elites play by their own rules, and that regular citizens are shut out of the real decisions about how powerful AI will shape jobs, privacy, and security.