Summary of Key Points
The U.S. government has required Anthropic to restrict access to its flagship models based on nationality, leading Anthropic to simply shut down its global services due to the complexity of implementing this policy. This incident exposes the contradictions in the U.S. approach to AI regulation and reflects the fundamental differences between China and the United States in their AI development paths. The U.S. follows a hegemonic route of "closed-source monopolies and policy control," while China adopts a resilient approach of "open-source collaboration and sovereign customization." The clash between these two approaches is shaping a new digital order known as "Digital Westphalia"—one based on national sovereignty, supported by open interoperability, and tangible assets such as energy and computing power, aiming to replace the U.S.-dominated technological hegemony.
I. The Anthropic Ban: A Self-Inflicted Blunder in U.S. Regulation
The U.S.'s directive to Anthropic was straightforward: foreign individuals (including its own international employees) were not allowed to use the Fable5 and Mythos5 models. However, Anthropic responded by stating that it would be impossible to verify users' nationalities effectively, and attempting to enforce this rule would only lead to chaos, so they chose to shut down the models for all users globally.
The issues with this ban are as follows:
- Flawed Justification: The ban was initiated due to a "escape room demonstration" (simulating a hacker attack to find vulnerabilities), but other models already possess similar capabilities. The U.S.'s directive lacks both transparency and technical justification.
- Practical Infeasibility: How can the global user base be accurately identified by nationality? It would be impractical to require everyone to upload their passports, leading to a blanket shutdown that affects hundreds of millions of users.
- Internal Frictions: Anthropic previously had a close relationship with the U.S. government (having conducted security tests), but it could not tolerate such opaque policy interference. This indicates that U.S. regulation has become so extreme that it disregards the actual needs of the company.
II. The AI Development Paths of China and the United States: Closed-Source Monopoly vs. Open-Source Sovereignty Customization
The paths taken by China and the U.S. in AI development are completely different:
- U.S. Approach: The U.S. treats advanced AI models as weapons, controlling them through closed-source practices (not sharing model code) to maintain a monopoly and using export controls to slow down global technology diffusion. For example, the 2025 "AI Diffusion Export Control Framework" designates cloud service providers (AWS, Azure) as gatekeepers, allowing only "trusted allies" to access these technologies. This strategy relies on mutual dependence as a weapon to compel other countries to rely on the U.S. However, this approach is becoming ineffective. The advantages of closed-source models (such as exclusive performance and data dominance) are being eroded by open-source models. For instance, China's DeepSeek R1 model has achieved similar reasoning capabilities to the U.S.'s o1 model at a lower cost and has become widely used, even among American developers.
- Chinese Approach: China focuses on "open-source collaboration and sovereign customization." Its GLM-5.2 model is released under the MIT License (one of the most permissive open-source licenses), and DeepSeek continuously updates its models. Developers can customize them according to local languages, regulations, and use cases without being constrained by the U.S.
The benefits of this approach include:
- Breaking Monopolies: Companies do not have to pay high API fees and are not at risk of being held hostage by platforms.
- Local Adaptability: Countries like Africa can use open-source models for local language translations, and Southeast Asia can develop agricultural AI without relying on U.S.-owned models.
- Cost-Effectiveness: Chinese models offer better performance and efficiency, making them suitable for deployment in various environments with lower geopolitical risks.
III. Digital Westphalia: A New Digital Order in a Multi-Polar World
The concept of "Digital Westphalia" draws from the 1648 Treaty of Westphalia, which established national territorial sovereignty. The core idea is that each country maintains sovereignty in the digital realm (data, standards, infrastructure) while using open protocols to ensure connectivity. This new order is characterized by:
1. Digital Sovereignty: Countries can set their own digital rules; for example, Chinese open-source models allow them to customize their AI systems, whereas U.S. bans impose its own rules on the global community.
2. Open-Source Interoperability: Similar to the Linux ecosystem, open-source code enables free modification and prevents reliance on dominant companies.
3. Collective Consensus: Local AI systems align with national values, avoiding fragmentation into isolated "information tribes."
4. Distributed Security: Cooperation is possible without relying on a central authority (e.g., U.S. cloud providers) through technologies like blockchain.
5. Public Data: Data is used for national governance and public welfare, not just for profit.
The Anthropic ban reflects the U.S.'s resistance to this new order; it seeks to maintain centralized control, while China's open-source approach contributes to its realization.
IV. The Foundation of a New Order: Energy and Computing Power
A digital world based on Digital Westphalia requires tangible foundations:
- Energy Advantages: AI training and inference are energy-intensive. China generates twice as much electricity as the U.S., with a diverse mix of renewable (hydro, wind), coal, and nuclear power sources, and lower industrial electricity costs. The "East Data, West Computing" policy has positioned computing centers in the western regions, reducing costs and promoting sustainability.
- Edge Computing Technologies: Advances in 2D semiconductors, graphene, and frictional nanogenerators (TENGs) enable low-power, self-sustaining AI systems that can process data locally, reducing reliance on overseas supply chains.
- Positive Cycle: Abundant clean energy drives more computing power, which in turn optimizes the energy and industrial systems, creating a virtuous cycle.
These strengths allow Chinese open-source models to be adopted globally, especially in developing countries with limited resources, such as Africa, where edge AI can be used for agricultural monitoring without relying on U.S. cloud services.
V. The Future Direction: How the Open Path Can Gain the Upper Hand
The outcome of this AI competition is no longer limited to two companies but represents a global choice regarding digital order:
- U.S. Dilemma: Closed-source models increasingly depend on policy protection (e.g., export controls), but open-source models are eroding their market share. If the U.S. loses its technological lead, the high profits that support OpenAI and Anthropic's valuations will collapse.
- China's Opportunities: The open-source approach not only empowers other countries but also allows China to build local intelligence capabilities and expand its technological and industrial landscape. Models like GLM-5.2 and DeepSeek have gained popularity among developers worldwide, proving the viability of this path.
- Global Impact: For developing countries, Digital Westphalia offers the opportunity for technological autonomy, enabling them to customize their AI systems independently from the U.S. The U.S.'s closed approach will only accelerate global technology fragmentation.
In summary, the future digital world will not be dominated by a single entity but by a multi-polar order where countries with open-source collaboration and tangible assets gain the upper hand. Those who invest in openness, sovereignty, and technological strength will have the advantage in this new era.
In one sentence: The U.S.'s attempt to control AI technology has backfired, while China's open-source approach is empowering the world and paving the way for a digital era of digital sovereignty. In this competition, openness and resilience will ultimately triumph over closure and hegemony.