The United States is considering a sharper test of technological alignment in its competition with China, with a draft State Department proposal warning countries that participation in a U.S.-led AI cooperation framework could be incompatible with membership in a rival China-backed initiative.
Reuters reported the proposal after reviewing a draft State Department letter. The document could still be revised, and the State Department has not announced the approach as formal policy.
If implemented, however, it would mark a significant evolution in the U.S.-China technology competition. The rivalry is moving beyond restrictions on advanced semiconductors into the broader infrastructure underpinning artificial intelligence — including critical minerals, computing capacity, energy, data centers, models, training data and technical standards.
For governments maintaining relationships with both Washington and Beijing, decisions once treated largely as technology or investment policy could become increasingly difficult to separate from foreign policy.
From chip controls to AI alliances
At the center of Washington’s strategy is Pax Silica, a U.S.-led initiative designed to strengthen trusted supply chains for AI, advanced semiconductors and critical minerals.
The initiative expanded to 24 signatories following a June 2026 summit. Participants include major U.S. allies such as Japan, Australia and South Korea, alongside countries including Kazakhstan.
The draft U.S. communication would warn governments against simultaneously joining competing technology initiatives associated with China.
Kazakhstan illustrates the tension. It participates in Pax Silica but has also joined China’s World Artificial Intelligence Cooperation Organization.
According to the draft reported by Reuters, Washington’s concern is that participation in competing frameworks could undermine the trust required for deeper cooperation on strategically sensitive technologies.
That distinction matters. The proposal does not necessarily prohibit governments from buying Chinese AI systems or doing business with Chinese technology companies. Rather, it would seek to limit simultaneous participation in competing government-led strategic partnerships.
China builds an alternative AI ecosystem
As Washington deepens its technology partnerships, Beijing is constructing an international AI architecture of its own.
China used the World Artificial Intelligence Conference in Shanghai in July to promote broader access to AI, open-weight technology and international cooperation. A new World Artificial Intelligence Cooperation Organization brought together representatives from 29 countries.
Chinese developers have also expanded the availability of capable open-weight models, creating another route for Beijing to increase its technological reach abroad.
But China’s strategy increasingly extends beyond exporting models.
Beijing also wants to become a major supplier of the data used to train them.
The New York Times reported this week that China is developing large, government-backed collections of training data while seeking to make Chinese datasets more widely available internationally. Earlier this year, China’s National Data Administration outlined plans to establish the country as a global data power by the end of 2028, including high-quality datasets across more than 20 strategic fields.
China has also pledged to share datasets with developing countries seeking to build their own AI systems.
The strategy addresses a technical challenge: China generates enormous volumes of information, but much of it remains fragmented across government agencies and companies. Beijing is seeking to break down those data silos while expanding the supply of higher-value information needed to train more sophisticated AI systems.
The Implications go beyond model performance
Training data helps shape what AI systems know, how they interpret questions and which linguistic, historical and cultural perspectives they reproduce. Chinese policymakers have argued that models trained disproportionately on English-language and Western data can misunderstand China or reflect Western perspectives on sensitive issues.
Critics, in turn, warn that reliance on Chinese government-produced datasets could increase the presence of state-approved narratives in AI systems deployed elsewhere.
The broader point applies to every AI ecosystem: who supplies the data increasingly matters alongside who supplies the chips and models.
The strategic dilemma for third countries
That widening competition creates a difficult calculation for governments with important relationships on both sides.
The European Union is pursuing greater technological sovereignty while remaining deeply integrated with U.S. technology companies and maintaining substantial commercial ties with China. Gulf states are investing heavily in AI infrastructure while cultivating relationships with both Washington and Beijing. Southeast Asian economies are attracting data-center and semiconductor investment from multiple partners.
Many governments may prefer diversified relationships rather than exclusive alignment.
Washington’s emerging approach, however, suggests that participation in strategically sensitive technology coalitions could increasingly come with conditions.
Closer alignment with the United States could improve access to advanced computing technology, investment and trusted supply chains. China can offer a different proposition: lower-cost and open-weight models, manufacturing capacity, infrastructure relationships and, increasingly, datasets and technical cooperation.
For governments seeking to preserve relationships with both, maintaining that balance may become harder.
AI competition expands into standards — and data
The rivalry is also extending beyond hardware and software into the rules and information systems governing how artificial intelligence is built and deployed.
AI ecosystems increasingly encompass cybersecurity standards, data governance, interoperability, model deployment, training datasets and infrastructure security.
Once those systems become embedded in national economies, switching technologies, suppliers or data architectures can become expensive and politically complicated.
That gives decisions being made today potentially long-term consequences.
Neither side is guaranteed to divide the world neatly into two camps. Europe, India, Gulf states and other middle powers have strong incentives to develop their own capabilities and avoid excessive dependence on either technological power.
But the more access to chips, models, datasets and infrastructure becomes tied to geopolitical alignment, the harder that autonomy may be to preserve.
What governments should watch
Access becomes leverage. Advanced chips, computing infrastructure and strategic technology partnerships could increasingly be offered through trusted geopolitical networks rather than purely commercial relationships.
Data becomes strategic infrastructure. Governments increasingly need to evaluate not only where their AI models come from, but which datasets train them, who controls those datasets and what dependencies they may create.
Procurement becomes foreign policy. Choices involving cloud providers, semiconductor suppliers, foundational models, data centers and training datasets may increasingly be interpreted as indicators of strategic alignment.
The Global South becomes a major arena of competition. China’s offer to provide models, training data and technical assistance could give developing countries alternatives to Western technology ecosystems — while raising new questions about standards, governance and dependence.
Middle powers may resist binary alignment. Governments with important relationships with both Washington and Beijing are likely to keep pursuing diversified suppliers and greater technological autonomy where possible.
Strategic Outlook
Washington’s proposed approach points toward a world in which access to strategic technology increasingly depends on geopolitical trust.
China’s expanding data strategy shows how quickly the competition is widening. The first phase of the AI race centered on compute — chips, data centers and capital. The next is increasingly about distribution — whose models, cloud systems and standards become embedded globally. A third layer is now emerging: information itself.
The countries and institutions supplying AI training data may influence not only which technologies succeed, but how those systems understand languages, societies and political questions.
That gives governments outside the United States and China new leverage. Countries controlling critical minerals, energy, semiconductor capacity, valuable datasets or fast-growing digital markets may find themselves increasingly courted by both sides.
But those choices can also create long-term dependencies.
The outcome is unlikely to be a clean division into two camps. The more Washington or Beijing conditions access on alignment, however, the harder it becomes for governments to preserve relationships with both.
The next phase of the AI race may therefore be determined not simply by who builds the most powerful technology, but by who builds the most attractive ecosystem around it.
For governments being courted by both sides, the strategic question is increasingly clear: whose chips, models and data they are prepared to depend on — and how much autonomy they are willing to trade for access.
RELATED STORIES:
Follow SDG News on LinkedIn
