US Plans to Give AI Partners a 'Multiple Choice': Pax Silica is Transitioning from a Collaborative Framework to a Technological Camp

CoinWorld reported:

According to a Reuters report on August 15, the US government is preparing to present clearer camp requirements to dozens of countries: if a country joins the China-led competitive AI cooperation framework, it may be excluded from the US-led Pax Silica alliance. The report is based on a statement from a US official and an internal draft seen by Reuters, so the accurate expression at this point is "the US is preparing" rather than "the new rules are in effect." It remains to be seen whether the draft will be implemented as is, which countries will be formally notified, and how exit or exemption mechanisms will be designed, pending public documents.

Pax Silica is not just a forum for discussing model safety. The US launched this initiative at the end of 2025, covering cooperation across cutting-edge models, semiconductors, advanced manufacturing, network infrastructure, and critical minerals. In June 2026, the government of Kazakhstan announced its membership, stating that the alliance then included partners such as Australia, India, Japan, South Korea, Singapore, the UK, and the UAE. The US aims to connect the chips, energy, minerals, data centers, and application markets needed for training models into a "trusted supply chain."

The change comes after China promoted another set of international AI cooperation arrangements. Reuters reported that Kazakhstan's simultaneous participation in both arrangements has raised concerns in Washington. If the draft ultimately becomes policy, countries that could have maintained flexibility across different projects will face a harder choice: not only deciding which model or chip to procure but also determining which set of rules to embed deeper into talent, data, critical materials, and infrastructure.

This issue cannot be simplified to "countries choosing between the US model and the Chinese model." Many countries simultaneously use US cloud platforms, Chinese open-source weights, European regulatory frameworks, and their own data centers. The truly difficult aspect to cut is the supply chain: minerals may be mined in one country, refined in another, manufactured in a third, and finally enter a data center in a fourth. An administrative multiple-choice question, when faced with commercial multi-layer dependencies, will have high execution costs.

AI competition is expanding from model rankings to the entire supply chain

In the past two years, the most observable indicator of AI competition has been model capability: whose reasoning is stronger, whose context is longer, and whose price is lower. But models are just one layer of the final product. Training and reasoning rely on advanced chips, which depend on manufacturing equipment, packaging, storage, and power, while data centers require permits, land, fiber optics, and stable power grids. Pax Silica incorporates these links into the same framework, indicating that the US is attempting to translate technological leadership into sustainable institutional and supply advantages.

"Exclusivity" is the new variable most worthy of attention in this report. Collaborative alliances typically attract members through joint investment and market access; if joining a competitive framework leads to exclusion, the alliance shifts from incremental cooperation to camp management with thresholds. For resource-rich countries, critical minerals may exchange for US capital and technology; for manufacturing countries, advanced equipment and orders may be more important; for larger market countries, considerations will weigh price, digital sovereignty, and long-term bargaining power. Different countries will not arrive at the same answer.

The challenge for the US lies in the fact that the competitiveness of Chinese AI products does not solely stem from national relations. Some Chinese models adopt open weights and offer lower deployment costs, allowing businesses to operate locally and reduce dependence on a single cloud service provider. If the US requires partners to abandon these options, it must provide sufficient alternatives, including affordable computing power, financing, talent training, and localized deployment. Merely relying on restrictions may force partners to declare their positions but may not ensure that actual technology usage migrates in sync.

Businesses will also be drawn into the policy boundaries. Multinational cloud vendors will need to reassess their customers and data center locations, chip and equipment manufacturers will need to determine end-use, and model companies may be required to provide clearer supply chain proofs. Compliance targets will no longer be just a batch of restricted chips but whether partners touch another ecosystem. The broader the audit scope, the higher the risk of inadvertently harming normal commercial cooperation and slowing project delivery.

Whether "picking sides" can be implemented depends on what the US offers, not just what it prohibits

From the US perspective, reducing the critical supply chain's exploitation by competitors has a clear national security logic. Cutting-edge AI may impact military planning, cyber offense and defense, and economic productivity, while critical minerals and advanced manufacturing capabilities are difficult to replicate in the short term. Fixing resource and technology partners within the same network can reduce the risk of supply disruptions and make export controls easier to coordinate.

However, for the alliance to be stable, members must see quantifiable benefits. Whether data center investments are timely, whether local businesses can access models and computing power, whether technical training creates jobs, and whether critical minerals have long-term procurement contracts are all more decisive than diplomatic slogans. If the US demands exclusivity but cannot guarantee financing, market, and technology supply, some countries may align on paper while continuing to maintain multilateral relationships in actual projects.

It is also important to note that the policy has not yet been formally announced. The internal draft may change after departmental coordination, ally feedback, and legal review, and the definition of "joining" may expand or narrow from formal signing to specific projects. For market participants, now is the time for scenario analysis, not for treating all cross-border AI cooperation as already prohibited.

What this news truly reveals is that the globalization of AI is entering a phase of higher friction. Models will still be called across borders, and open-source code is difficult to completely seal off by geographical boundaries, but chips, capital, minerals, and large data centers are more easily controlled by policy. The next stage of AI competition will not only be determined by who releases the strongest model first but also by who can keep more countries in their technological network with more attractive, executable, and stable conditions.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.