Anthropic vs. China: The Secret AI War Exploding Behind Your Screen

Anthropic’s quiet deployment of tracking code inside its Claude Code chatbot in March exposed a growing and increasingly aggressive technological conflict between U.S. and Chinese AI companies. The code, which checked whether users were operating in Chinese time zones or using domains linked to Chinese AI labs, was intended to identify entities Anthropic suspected of “distilling” its models…a process where a powerful AI is used to train a smaller, cheaper one by generating massive amounts of responses. Distillation is legal and widely used across the industry, but doing so without permission violates company rules and, according to Anthropic and OpenAI, is enabling Chinese competitors to rapidly close the performance gap. 

Anthropic removed the tracking code after public criticism, calling it an experiment, but the revelation highlighted the escalating stakes. Chinese AI companies have repeatedly matched U.S. model capabilities within months, and free Chinese models are becoming more popular globally. Cybersecurity firm Semgrep recently found that Zhipu AI’s free model outperformed Anthropic’s Claude Opus 4.8 in vulnerability detection. 

Anthropic alleges that Chinese firms…including Alibaba’s Qwen team, Deepseek, Moonshot, and MiniMax…have used tens of thousands of fraudulent accounts to generate tens of millions of interactions with Claude, effectively siphoning its intelligence. A February 2025 study by Peking University and the Chinese Academy of Sciences found strong evidence that many Chinese models, including Alibaba’s Qwen, mimic Claude’s behavior, suggesting distillation at scale. 

The U.S. government has echoed Anthropic’s warnings, framing distillation as a national security threat that could allow China to build cheaper models stripped of safety features. The Trump administration has tightened export controls and pressured AI companies to restrict access, but Silicon Valley leaders argue that overly strict policies may hinder U.S. competitiveness. 

Despite restrictions, Chinese users routinely bypass blocks using proxy networks, resold accounts, and verification workarounds. Anthropic has banned nearly 700,000 China-based accounts, yet access to Claude and OpenAI models can still be purchased for as little as $1 per month on Chinese marketplaces. 

Experts caution that labeling Chinese models as “stolen” risks underestimating China’s genuine innovation, especially in efficiency and cost reduction. Meanwhile, U.S. startups and even Fortune 500 companies increasingly rely on Chinese open-source models due to their affordability. The conflict underscores a broader geopolitical struggle over who will control the future of AI…and whether the U.S. can maintain a technological lead. 

Original article found here: https://www.washingtonpost.com/national-security/2026/07/06/why-anthropic-alleges-chinese-firms-are-distilling-knowledge-claude/ 

*****

Get Ageless Tech. Using AI to Your Advantage: https://georgemagazine.com/agelesstech/

Get George’s America’s 250th: https://georgemagazine.com/product/america-250-celebrating-legacy/

Get George’s Crypto E-Playbook: https://georgemagazine.com/product/the-crypto-playbook-by-george-magazine/

Get America’s 250th George Magazine T-Shirt: https://georgemagazine.com/product/americas-250th-george-magazine-t-shirt/

Get The George Dispatch Free: https://georgemagazine.com/thedispatch/

Get George Magazine. Print or Digital: https://georgemagazine.com/subscribe-george-magazine/

Get George Junior. Print or Digital: https://georgemagazine.com/subscribe-george-junior-magazine/

Get the Aetherian Realm’s Final Judgments Map: https://georgemagazine.com/product/aetherian-realms-final-judgments-map/

Go to Project Looking Glass: https://projectlookingglass.org/