Morning Edition · Friday, June 26, 2026
US-Led Pax Silica Bloc Widens to 24 States as AI Supply Chains Split Along Geopolitical Lines
European Union members joined the American initiative to govern AI hardware and model supply, weeks after Washington ordered access to two Anthropic frontier models suspended under export controls.
The United States initiative known as Pax Silica, which seeks to coordinate allied control over the hardware and the model supply behind artificial intelligence (AI), has grown to a reported 24 participating states after European Union members joined, according to a summary circulated by the Russian-language channel AI ML Big Data. The same account lists Kazakhstan, Argentina, Chile, Costa Rica, and Panama as the next prospective members. These figures come from a single secondary post and have not been confirmed by an official government statement, so the exact count should be treated as asserted rather than independently verified.
The direction is consistent with policy already on the record. Earlier this month, Anthropic confirmed that the United States government issued an export-control directive ordering it to suspend all access to two of its models, Fable 5 and Mythos 5. That action extended export controls beyond chips to the models themselves, treating model weights and inference access as restricted strategic goods.
The combination matters because it points to a coordinated framework rather than isolated measures. Bringing European partners and several Latin American and Central Asian states into a single arrangement would formalize a bloc that aligns chip supply, model access, and end-use rules. States outside such an arrangement have a clear incentive to standardize on downloadable Chinese open-weight models, which carry no comparable access restriction.
If the claims hold, the beneficiary is clear. A larger Pax Silica strengthens Washington's leverage over the spread of frontier capability and over the hardware that trains it. It also raises the cost for any government deciding whether to build on American or Chinese systems, which is the practical mechanism by which a single global frontier splits into competing supply chains.
- If true, who benefits
Washington and allied chip and frontier-model incumbents, who gain regulatory leverage over which states can buy compute and model access, while excluded buyers are pushed toward cheaper Chinese open-weight systems.
- The nuance
The Telegram count is real, multiple outlets confirm the EU joining and a roughly 24-state membership, but Pax Silica is a non-binding declaration, not an enforceable bloc, and China frames it as a coercive "small clique" rather than neutral supply-chain governance.
An open-source-intelligence read of how likely this story is true with its real nuance, not a judgment of any outlet. It assesses the claim, weighing independent and adversarial reporting. How we label confidence.
What this means
Export control is moving from the physical layer (chips and lithography tools) to the logical layer (model weights and application programming interface, or API, access). For engineers and operators, that means the legal availability of a given frontier model can now depend on the customer's jurisdiction. Multi-vendor and open-weight fallback options become a procurement requirement rather than a preference.
What to watch
- Whether any government publishes an official Pax Silica membership list and binding rules, which would turn a reported alliance into enforceable policy with concrete compliance obligations.
- Whether more labs disclose government-ordered suspensions of specific models, which would show access restriction becoming a routine instrument rather than a one-time event.
- Adoption rates of Chinese open-weight models in states excluded from the bloc, the clearest sign that the world is standardizing on two systems instead of one.
Observations to monitor, not financial advice.
Synthesized from: Polylog editors · Anthropic News
Part of a tracked trend
AI Sovereignty and Export Controls on Frontier Models
Over the next 3-6 months, governments increasingly treat frontier AI models as strategic national assets — extending export controls to model access itself and backing domestic 'champion' labs as sovereignty plays.
More from this edition
- Agents Become the First Real Economic Test of the AI Buildout
- IBM Discloses 0.7-Nanometer Transistor Process Using Vertically Stacked Devices
- MindOn Shows a Single Model Controlling Robots of Different Body Types
- Unitree Prices a Humanoid Robot From 4,900 Dollars
- New Interpretability Work Locates Sycophancy and Refusal in the Same Activation Geometry
- Papers Ask What to Do After a Benchmark Saturates and How to Tell Knowing From Guessing
- A New Benchmark Tests Whether AI Agents Can Actually Secure Smart Contracts
- Study Finds Standard Post-Training Can Erode Values Instilled During Pretraining
- Anthropic Pushes Policy Proposals Built Around an AI Capability Curve It Calls Exponential
- Survey Data Show Gen Z Both the Heaviest AI Users and the Most Skeptical
- Researchers Propose Turning Editing Models' Own Moderation Filters Into a Defense