Andrew's Notes

When China's Open-Source AI Is a Trap

Properties5
authorThe Economist
descriptionA Telegram column in The Economist arguing that China's promotion of open-source AI is a strategy of the laggard rather than an act of benevolence, that Chinese models carry political-compliance requirements, and that Beijing would abandon openness the moment it collided with national security or political power.
published2026-07-14
sourcehttps://www.economist.com/international/2026/07/14/when-chinas-open-source-ai-is-a-trap
tagsai-policy, open-weight-models, china, us-china, ai-governance, geopolitics, censorship

4 min read

When China’s Open-Source AI Is a Trap

Abstract

China’s presentation of open-source AI as a gift to humanity is strategy, not benevolence: openness is the rational play for a laggard whose top models still trail the best American ones. Countries fearing dependence on a control-obsessed America should not pin their hopes on China instead — Chinese models come with political-compliance obligations, Beijing is already reviewing its commitment to openness, and if openness ever clashes with national security or political power, control wins instantly.

Framing: the WAIC pitch

  • Column pegged to the World Artificial Intelligence Conference (WAIC) opening in Shanghai on 17 July, where Xi Jinping was due to speak.
  • Party media previews trailed phrases such as “those who walk together go far” and “global AI for good”.
  • In China’s telling, benevolence explains why its large language models are open-source or open-weight — models users can download, run on their own servers and customise.
  • China describes open-source AI as a shared asset for all humanity, aimed notably at users in less wealthy countries.

The governance pitch

  • Xi was expected to argue the UN should lead a push for global AI governance, with China offering a dose of “Chinese wisdom” as new rules are crafted.
  • In policy forums and papers, China promotes “inclusive” AI rules that respect different political systems and the sovereign rights of states.
  • The contrast with America is deliberate. Successive US presidents and democratic allies have talked of embedding liberal values and freedoms in AI; China paints that as Western chauvinism.
  • What China offers instead: non-judgmental AI for economic development, with no questions about censorship or how foreign leaders gain and keep power.

Why the pitch finds an audience

  • These are jarring times for users of American AI technologies: the Trump administration has recently revoked access to powerful AI tools where it judged controls necessary for national security or to maintain “AI dominance”.
  • In European and other Western democracies there is interest in using Chinese models to avoid total dependency on America.

Why China is not the safe alternative

Political compliance is built in

  • Strict rules require Chinese AI firms to uphold national security, social stability and “core socialist values”.
  • Cyber-regulators test LLMs, bots and agents for political compliance, bombarding them with tricky questions.
  • Illustrative case: American researchers asked Miiloo, an AI-enabled doll exported from China, about the status of Taiwan; the toy replied that the island is an inalienable part of China and that this cannot be refuted.

A record of using industrial might for leverage

  • Precedent cited: China curbed exports of Chinese-processed rare-earth elements and magnets last year, bringing Trump to heel.
  • China’s AI ambition is to sell the world an ecosystem — Chinese-made models and tools, computer chips, data centres and cooling systems, ideally powered by Chinese-made electrical grids and renewable energy plants.

Openness is a laggard’s strategy

  • Chinese officials present open-source AI as a gift; in reality openness is a logical strategy for laggards.
  • The performance of China’s top models remains some way behind the best American LLMs, which makes it rational to woo foreign customers with cheaper, downloadable models as an alternative to expensive proprietary American ones.
  • Domestically, state planners want companies to build clever AI applications to unleash a productivity revolution and a consumption boom — cheap open-source tools help.

Signs Beijing is already reconsidering openness

  • Officials dread foreigners swiping tech secrets.
  • In April, Chinese regulators ordered Meta to unwind its purchase of Manus, an AI-agent startup (despite the Chinese co-founders having moved Manus to Singapore). China has since tightened rules on all cross-border AI deals.
  • Reuters reported discussions between Chinese regulators and companies about possibly limiting foreigners’ access to China’s most advanced models.
  • AI safety is now a live concern: Chinese military figures reportedly expressed shock at how America and Israel used AI to find targets in Iran. If China develops near-Mythos-grade tools capable of building destructive cyber- or bioweapons, its leaders have cause to fear giving adversaries access.

Intensifying competition

  • Rumours that the Trump administration is weighing whether to stop Americans using Chinese AI models, at least for government work.
  • Two explanations offered: national security, and the growing popularity of Chinese LLMs as a cheap tool for basic tasks.

Conclusion

  • Embracing China is a risky hedge against a domineering America.
  • China has always been an unlikely champion of openness; Party chiefs enjoy the propaganda win of painting America as a bully, and hope low-cost AI will hook foreigners on Chinese digital infrastructure.
  • If openness ever clashes with national security or political power, they will choose control in an instant.