Why Are We Silently Watching China’s AI Invasion?
Recently, Airbnb’s CEO Brian Chesky acknowledged that his company is “very dependent” on Chinese AI models due to their efficiency and lower costs. This should have raised concerns in Washington. It’s telling that he’s opting for Alibaba’s Qwen over his acquaintance Sam Altman’s OpenAI—not just for quality, but mainly because the Chinese option is significantly cheaper.
Bloomberg columnist Katherine Sobecke pointed out that Alibaba’s Qwen has racked up approximately 385 million downloads, surpassing Meta’s Llama, which has around 346 million. Currently, Chinese AI models represent over 40% of the new language models being released, while Meta’s share sits at about 15%. Additionally, US-based coding startups valued at around $10 billion, including Cursor AI and Cognition AI, seem to be operating on Chinese AI infrastructure, although these companies haven’t confirmed this.
Meanwhile, accessing ChatGPT or Claude from China is impossible. According to state media, the Cyberspace Administration of China has approved over 100 domestic AI models for public access, leaving US products effectively shut out of the market. This is due to a combination of stringent regulations and the Great Firewall.
This represents a clear case of trade imbalance: China has fortified its barriers against American AI companies while giving Chinese services free rein to access American users and developers. And we seem to be enabling this situation.
In response, Washington has proposed several measures, but they have largely missed the mark. The Adversarial AI Act aims to prevent US government agencies from using Chinese AI, and Sen. Josh Hawley’s bill seeks to criminalize American work on Chinese AI research. While there is merit to many of these suggestions, they mostly target government dealings or discourage investments. However, they overlook a core issue: American companies can buy Chinese AI services freely, while Chinese firms aren’t allowed to procure American AI services—even if they wanted to. This one-sided barrier contradicts the promises made by President Donald Trump.
Moderate Reciprocity in AI
So why is action necessary? First, there’s the principle of reciprocity. If China restricts American AI from entering its market, we should consider blocking Chinese AI in return, or at least implementing tariffs.
Secondly, Chinese AI companies benefit from substantial state subsidies, enabling them to undercut US competitors. As venture capitalist Chamath Palihapitiya noted, Chinese AI is “frankly a ton cheaper” than options from OpenAI or Anthropic. This price discrepancy arises not from market forces, but from China’s industrial policy. Tariffs could help level the playing field by requiring US buyers to account for these hidden subsidies.
Thirdly, there are legitimate national security concerns about what US companies are supporting within China’s AI systems (which the Chinese government can access due to local laws). Imposing tariffs could require companies to address security risks rather than passively accepting them.
It’s vital to recognize that true reciprocity may entail a complete ban on Chinese AI, given that China is not only restricting but outright banning American AI. An immediate step could involve a more measured approach: applying tariffs on Chinese AI services. This has multiple benefits. It generates revenue rather than simply halting commerce, allowing market dynamics to function. If the price remains attractive even after tariffs, companies can still opt for Chinese AI.
The process would be straightforward. When US companies pay for Chinese AI services through various means, such as API calls and commercial licenses, these transactions would be taxable. Enforcement could follow existing frameworks for other imported digital services.
Critics might argue that this would boost profits for American companies, leading to “globally low competitiveness.” This line has persisted over many years as we let our industrial base slip to China. Allowing this to occur again—this time in AI—would be a severe misstep.
The pressing question is whether we’re genuinely committed to defending America’s AI leadership. We can’t afford to give Chinese AI companies unrestricted access to American consumers while complaining about China’s industrial policies. It’s time we practice what we preach regarding reciprocity and impose tariffs on Chinese AI.
