China wants the world to use its AI – but not all of it

At the World Artificial Intelligence Conference in Shanghai on July 17, Chinese President Xi Jinping set out China’s clearest vision yet for the future of artificial intelligence.

Beijing would not merely build world-class AI. It would help other countries develop it, promote Chinese open-source models, train thousands of AI practitioners across the developing world and build institutions to shape global AI governance.

Yet, even as Xi was urging wider international cooperation, Chinese regulators were reportedly discussing whether overseas access to the country’s most advanced AI models should be restricted.

Beijing is discovering that technological leadership requires managing two competing objectives: keeping its most valuable technologies under control while making Chinese technology indispensable to the rest of the world.

There are compelling commercial reasons for this. China’s AI companies are becoming globally competitive, while its semiconductor and computing equipment exports have surged alongside the worldwide AI investment boom.

In June, its semiconductor exports jumped 122 per cent by value from a year earlier and computing equipment exports rose 53 per cent. Expanding overseas adoption of Chinese models promises not only greater geopolitical influence but also new export markets for Chinese chips, cloud services and AI infrastructure.

That outward-looking vision, however, is only one side of Beijing’s strategy. As it seeks to spread Chinese AI around the world, it is also tightening control over the technologies it considers most strategically valuable.

Oversight, controls and scrutiny

On July 1, it significantly strengthened its oversight of overseas transactions involving Chinese technology, data and national security with beefed-up regulations that give Beijing broad powers to unwind completed deals, block unauthorised transfers of sensitive technologies and scrutinise investments that could undermine its strategic interests.

It also established a reporting mechanism encouraging organisations and individuals to flag suspected attempts to circumvent export controls, including routing sensitive technologies through third countries.

Meanwhile, Beijing has continued to expand its export control regime. It has introduced new control and watch lists restricting dual-use exports to designated foreign companies, including rare earth firms in the United States and technology companies in Japan.

On July 7, Reuters reported that Chinese regulators had met some of the country’s biggest tech companies, including Alibaba, whose flagship AI model is Qwen, and ByteDance, the developer of Doubao, to discuss whether overseas access to China’s most advanced AI models – both proprietary and open-weight – should be restricted.

Officials also reportedly raised the possibility of tightening scrutiny over who can invest in Chinese AI start-ups.

The discussions come as Washington has begun imposing its own controls on frontier AI. In June, the Trump administration barred foreign nationals from accessing Anthropic’s most advanced models, Fable and the cybersecurity-focused Mythos, over national security concerns.

While restrictions on Fable were later eased after new safeguards were introduced, Mythos remains available only to selected trusted American organisations.

Beijing has reportedly been watching these developments closely, concerned that frontier models such as Mythos could be used to identify software vulnerabilities or otherwise be deployed against Chinese interests.

It has begun tightening oversight of its own AI ecosystem, ordering Meta to unwind its US$2 billion (S$2.58 billion) acquisition of Chinese-founded, Singapore-based AI start-up Manus and investigating whether the company and other AI start-ups that relocated overseas may have breached China’s export-control laws.

A matter of strategic assets

If Beijing does decide to limit overseas access to its most advanced AI models, it would not represent a sudden departure so much as the logical extension of a strategy years in the making.

Under Xi, technology has been elevated from an engine of economic growth to a pillar of national strength. AI is no longer treated as an emerging industry but as a strategic technology and national resource.

China’s ambition is also increasingly institutional. At the conference in Shanghai, nearly 30 countries joined the newly launched World Artificial Intelligence Cooperation Organization, which Beijing hopes will become a platform for AI cooperation, standards and governance across the Global South.

China also released an action plan to promote its open-source AI models overseas and establish international AI cooperation centres.

These ambitions rest on China’s growing confidence that it is no longer merely catching up technologically, but is increasingly setting the pace.

According to think-tank the Australian Strategic Policy Institute’s latest Critical Technology Tracker, China now leads the world in high-impact research in 69 of 74 critical technologies, up from 66 in earlier assessments, spanning fields from AI and semiconductors to robotics, advanced materials and biotechnology.

China now poses a “high monopoly risk” in 41 technologies, meaning its lead is so substantial that catching up will become increasingly difficult.

It is also the world’s only country with a presence across all 41 major, 207 intermediate and 666 minor categories in the United Nations’ International Standard Industrial Classification. That gives it an unusually complete industrial base, spanning almost every stage of production and allowing it to manufacture a vast range of goods within its own borders.

But China is now confronting the same dilemma that faces the US. It wants its AI models to be adopted around the world, helping Chinese companies gain market share, expand Beijing’s technological influence and shape global standards.

Yet, it also sees its most advanced models as strategic assets that must remain secure and under Chinese control. The more successful those models become, the more difficult it is to reconcile openness with security.

From tech taker to keeper

For four decades, China’s priority was to draw technology in. From the start of reform and opening up, it courted foreign investment, used access to its vast market to encourage technology transfer and sent large numbers of students overseas to learn from the world’s leading universities and companies.

The aim was not simply to attract capital, but to absorb knowledge and build domestic capability.

By the 2000s, that strategy was giving way to a much heavier emphasis on indigenous innovation. Now that China is developing frontier technologies of its own, the direction of travel is changing again. Instead of focusing only on how to acquire advanced know-how, Beijing is increasingly concerned with how to keep strategically valuable technologies from flowing out too freely.

That does not necessarily make China hypocritical. Britain and the US also embraced openness once it suited their stage of development, while protecting technologies and industries they considered vital to national power.

Britain is remembered as the great champion of 19th-century free trade, especially after the repeal of the Corn Laws in 1846. But by then, it had already built the world’s leading industrial economy. For much of the 18th century, it had restricted the export of textile machinery and sought to prevent skilled craftsmen from emigrating, to stop rivals from copying its industrial methods.

The US followed a similar path. Throughout the 19th century, it maintained high tariffs to protect domestic manufacturing. After World War II, once it had become the dominant industrial and technological power, it had more to gain from opening foreign markets and became the principal architect of the post-war trading system, beginning with the General Agreement on Tariffs and Trade and later the establishment of the World Trade Organization.

Even then, openness had limits. During the Cold War, Washington restricted exports of advanced computers, semiconductors and other dual-use technologies to the Soviet bloc. In recent years, it has imposed sweeping controls on advanced AI chips, semiconductor manufacturing equipment and related technologies bound for China.

Beijing’s comprehensive industrial ecosystem, together with its dominance in areas such as rare earth processing and parts of the clean-energy supply chain, has given it considerable leverage over global manufacturing. But restricting access to frontier technologies could prove costly.

Those tighter export restrictions and heightened scrutiny of overseas technology deals are likely to accelerate efforts by the US and its allies to diversify supply chains and develop competing capabilities.

The repercussions are already being felt. The Financial Times reported on July 16 that the European Union is creating an emergency task force to prepare for a possible trade conflict with China when a temporary truce over rare earth export controls expires, while also drafting new laws to reduce dependence on single-country suppliers of critical inputs.

At home, more stringent controls could make it harder for Chinese firms to expand overseas, just as Beijing is trying to reap the commercial rewards of its technological advances.

China wants its AI to become part of the world’s digital infrastructure, and it now appears to be pursuing its own version of that ecosystem strategy – using affordable, open-source AI models, training programmes and new international institutions to embed Chinese technology across the developing world.

If successful, the benefits would be both commercial and geopolitical: more users, more developers, greater demand for Chinese AI infrastructure and a stronger voice in setting global standards.

But technological leadership also brings difficult choices; the technologies that generate the greatest economic and geopolitical influence are often the ones governments become most reluctant to share.