Redistributed

China pursues AI development through state direction rather than market forces. Understanding this alternative reveals something important: the US model of corporate-led AI development is a choice, not an inevitability. Different governance structures produce different distributions of value, risk, and power.

The Strategic Vision: AI as National Priority

In July 2017, China's State Council released the "New Generation Artificial Intelligence Development Plan" (AIDP). The document declared AI a national strategic priority and set explicit targets: achieve parity with global leaders by 2020, make major breakthroughs by 2025, become the world's primary AI innovation center by 2030 (State Council, 2017).

This was not aspiration masquerading as policy. The plan allocated substantial resources, designated responsible agencies, and integrated AI into the broader "Made in China 2025" industrial strategy. As Stanford's DigiChina project documented through translations and analysis, the Chinese government views AI not as one industry among many but as a general-purpose technology that will reshape economy, society, and military capability (DigiChina, 2017).

The logic is explicit in Chinese policy documents: AI represents a rare opportunity for technological leapfrogging. In semiconductors and software platforms, China remains behind. In AI, the argument goes, the field is young enough that catching up — and potentially surpassing — is achievable with sufficient investment and coordination.

The stakes, from Beijing's perspective: Whoever leads in AI gains advantage across domains — economic productivity, military capability, social governance, and geopolitical influence. This is not a market to be left to market forces.

State-Directed vs. Market-Driven Development

The contrast with the US model is structural, not just rhetorical.

US approach:

Private companies lead development (OpenAI, Google, Anthropic, Meta)
Government provides limited research funding, sets some guardrails
Capital markets allocate resources based on expected returns
Competition occurs between firms, with antitrust as the main check on concentration

Chinese approach:

State sets strategic direction and targets
Government directly invests in AI firms and research
National champions receive preferential treatment
Competition between firms occurs within state-defined parameters
Industrial policy coordinates across sectors

As Kai-Fu Lee documented in AI Superpowers, this creates different dynamics. The US model generates rapid innovation through competition but concentrates power in whoever wins the market. The Chinese model can mobilize resources at scale but risks the inefficiencies of top-down planning (Lee, 2018). The emergence of DeepSeek as a competitive frontier model, developed with creative efficiency under export control constraints, suggests that the Chinese model can produce genuine technical innovation, not just scaled-up imitation.

Neither model is inherently more "redistributive." Both concentrate power — just in different hands.

The US model concentrates power in:

Large tech companies and their shareholders
Venture capital and institutional investors
Talent pools in a few geographic hubs (Bay Area, Seattle, New York)
Whoever can pay for API access

The Chinese model concentrates power in:

The Chinese Communist Party and state apparatus
National champions with state backing
Central planning institutions
The surveillance and control capacities of the state

The question is not which model is better in the abstract. It's which model you live under, and what mechanisms exist to distribute the benefits and constrain the harms.

Key Players: National Champions and State Labs

China's AI ecosystem comprises a few dominant private companies operating with state support, state-owned enterprises, and national research institutions.

The "BAT" + Huawei

Baidu — search, autonomous vehicles, cloud computing. Baidu's Apollo autonomous driving platform represents the most visible Chinese effort in self-driving technology. The company operates AI research labs in China and the US (though the latter has faced increasing scrutiny).

Alibaba — e-commerce, cloud, logistics AI. Alibaba Cloud provides the compute infrastructure for much of Chinese AI development. The company's DAMO Academy conducts AI research across language, vision, and decision systems.

Tencent — social media, gaming, fintech. WeChat's ecosystem generates enormous behavioral data. Tencent's AI Lab focuses on applications within its platforms — content recommendation, translation, gaming AI.

Huawei — telecommunications, hardware, increasingly chips. After US sanctions cut off access to advanced semiconductors, Huawei accelerated its own chip development. The company's Ascend AI processors aim to reduce dependence on Nvidia, though they remain behind on performance (CSET, 2023).

State Institutions

Chinese Academy of Sciences — China's premier research institution, with AI research institutes across the country. CAS conducts foundational research and trains researchers who move to industry.

National Laboratory for Pattern Recognition — Focused on computer vision and image recognition, a capability with obvious surveillance applications.

Ministry of Science and Technology — Coordinates national AI research programs and funding. Designated "national AI innovation platforms" assign specific companies to lead in specific domains: Baidu for autonomous vehicles, Alibaba for smart cities, Tencent for medical imaging, iFlytek for speech recognition.

Military-civil fusion: Chinese policy explicitly encourages collaboration between civilian AI research and military applications. The boundaries between civilian and military AI development are intentionally blurred (Allen, 2019). It is worth noting that the US has its own version of this dynamic — DARPA funding of AI research, defense contracts with frontier labs, and the revolving door between Silicon Valley and the Pentagon — though the Chinese approach is more explicit and institutionally formalized.

Surveillance Applications: The Control Dimension

China's AI deployment includes extensive surveillance infrastructure. This is not incidental to Chinese AI strategy — it is central to it.

The "social credit system" encompasses multiple overlapping systems — government databases, commercial scoring, local pilots — rather than a single unified system. The concept: aggregate data on citizens' behavior to produce scores that determine access to services, credit, travel, and social standing.

Implementation varies by locality. Some versions resemble commercial credit scoring. Others incorporate ideological compliance. The opacity makes assessment difficult, which may be the point. As Rogier Creemers has documented through translations of primary sources, the system aims to make trustworthiness calculable and consequences for untrustworthiness automatic (Creemers, 2018).

From a redistribution perspective: Social credit redistributes power toward whoever defines "trustworthiness" and controls the scoring system. It extends state capacity into everyday transactions.

Facial Recognition

China has deployed facial recognition at scale: in subway systems, at building entrances, in classrooms to monitor student attention, in public spaces for law enforcement. Chinese companies — SenseTime, Megvii, Hikvision — lead globally in facial recognition deployment.

This creates a feedback loop: deployment generates training data that improves algorithms that enable more deployment. The scale of Chinese deployment provides competitive advantage in algorithm development.

Xinjiang

The most extreme application: mass surveillance of Uyghurs and other minorities in Xinjiang. Facial recognition, gait analysis, phone tracking, mandatory spyware, checkpoints with biometric capture — the infrastructure of a surveillance state deployed against an ethnic minority.

Human rights organizations and investigative journalists have documented forced labor, mass detention, and cultural suppression enabled by AI surveillance (Human Rights Watch, 2019; BuzzFeed News, 2020).

US export controls explicitly cite Xinjiang surveillance as justification for restricting semiconductor exports to Chinese AI companies. Several Chinese AI firms were placed on the Entity List specifically for their role in Xinjiang.

This is not separate from "AI development." The surveillance apparatus uses the same computer vision, natural language processing, and prediction systems that power commercial applications. The talent, data, and infrastructure flow between commercial and control applications.

Export Controls: The Chip War

Beginning in 2019 and accelerating in 2022, the US imposed export controls restricting China's access to advanced semiconductors and semiconductor manufacturing equipment.

The October 2022 rules were particularly aggressive. They blocked:

Advanced GPUs (Nvidia A100, H100, and equivalents)
Advanced chip manufacturing equipment, especially from ASML
US persons (citizens and green card holders) from working in Chinese semiconductor fabs

As Chris Miller documented in Chip War, this represents an attempt to freeze China's semiconductor industry at a particular technological level while the US and allies continue advancing (Miller, 2022).

The logic: AI training requires massive compute. Compute requires advanced chips. Control the chips, and you constrain AI capability development.

China's response has been threefold:

Stockpiling — Before restrictions took effect, Chinese companies acquired as many advanced chips as possible. Nvidia reported record China sales in 2023 as customers rushed to buy before cutoffs.
Substitution — Nvidia developed modified chips (A800, H800) with reduced capabilities that initially fell below export control thresholds. The US subsequently tightened rules to capture these variants too.
Indigenous development — Huawei's Ascend processors, SMIC's chipmaking efforts, and massive government investment aim to build domestic capability. Progress has been real but remains generations behind the frontier.

The redistribution lens: Export controls are a form of technology governance by other means. They redistribute capability from China to US allies. They also concentrate power in whoever controls the chokepoints — currently, ASML for lithography and TSMC for fabrication.

The chip war makes explicit what was implicit: AI capability depends on physical infrastructure that is geographically concentrated and politically contested.

Talent and Research: Publication Output and Brain Circulation

China's AI research output has grown dramatically. By several measures — total publications, citations, conference papers — Chinese institutions produce comparable volume to the US (CSET, 2024).

Quantity does not equal quality. Analysis by Georgetown's CSET suggests Chinese AI research is stronger in applications (computer vision, speech recognition) than in foundational architecture or theoretical work. The US retains advantages in breakthrough research (Zwetsloot et al., 2021).

Brain circulation complicates the picture:

For decades, Chinese students studied at US universities, then remained in the US to work. This "brain drain" benefited US AI development. The pattern is shifting. More Chinese-trained researchers return to China, attracted by opportunities, patriotic sentiment, or concern about treatment in the US.

US visa restrictions, particularly during the Trump administration and continuing in modified form, accelerated this shift. Treating Chinese students and researchers as potential threats pushed talent back to China.

The dynamics are recursive: Chinese-American researchers face suspicion in the US; some return to China; this is cited as evidence that Chinese-Americans are security risks. The "China Initiative" prosecuting academics for alleged China ties produced high-profile failures and was formally ended in 2022, but the chilling effect continues.

From a knowledge redistribution perspective: Talent flows shape whose institutions develop AI capability. US restrictions have made the global research community more fragmented.

Alternative Standards: Shaping Global AI Governance

China actively participates in international standards bodies, attempting to shape technical standards and governance frameworks.

Technical standards: At ITU, ISO, and other bodies, China has advocated for standards that favor Chinese technology approaches. Standards seem technical but carry economic stakes — whoever's approach becomes standard gains deployment advantages.

Governance frameworks: China has proposed AI governance principles that emphasize state sovereignty over data and AI systems, in contrast to US emphasis on private sector leadership or European emphasis on individual rights.

Belt and Road and AI: Chinese AI companies export surveillance and smart city technology to countries participating in the Belt and Road Initiative. This spreads not just technology but governance models. Countries adopting Chinese surveillance infrastructure become more integrated into Chinese technology standards and, potentially, Chinese political influence (Feldstein, 2019).

The governance question: As AI governance frameworks develop, who shapes them matters enormously. Different frameworks privilege different values: individual privacy, state control, commercial freedom, collective welfare. China's participation in governance discussions is an attempt to ensure frameworks that accommodate rather than constrain Chinese AI development and deployment.

What the Comparison Reveals About Choice

The US-China comparison is often framed as competition — who is "winning" the AI race. This framing obscures the more important question: what choices are available in how AI is developed and governed?

The Chinese model demonstrates that:

State direction of AI development is possible at scale
Industrial policy can mobilize resources and coordinate across sectors
AI can be deployed for social control as easily as for economic productivity
Technology transfer and talent flows are policy variables, not natural forces
Surveillance applications and commercial applications share infrastructure

The US model demonstrates that:

Market competition can produce rapid innovation
Private control also concentrates power, just in corporations rather than states
Weak governance creates vacuums that companies fill with self-regulation
AI capability depends on hardware controlled by a handful of companies
"Open" development coexists with massive concentration

Neither model is redistributive in the sense of spreading AI benefits broadly.

In China, the state captures value and directs it according to state priorities — which include mass surveillance, military capability, and social control alongside economic development. Citizens have limited voice in these priorities.

In the US, shareholders and executives capture value, and deployment follows profit opportunities. Workers, data subjects, and affected communities have limited voice.

Different pathways exist. European AI governance attempts to create rights-based constraints. Open-source communities attempt to build alternatives outside corporate or state control. Public AI initiatives attempt to direct development toward public purposes. None has yet achieved the scale of US corporate or Chinese state models.

The Redistribution Question

Ursula Franklin distinguished "holistic" technologies that enhance human capability from "prescriptive" technologies that reduce humans to following machine outputs (Franklin, 1989). Both US and Chinese AI development have tended toward prescriptive applications — algorithmic management of workers, automated decisions about credit and benefits, surveillance at scale.

The comparison reveals that AI governance is political, not technical. The question is not what AI "wants" to become but what interests shape its development and deployment.

For readers in liberal democracies, the Chinese model serves as:

A demonstration that alternatives to corporate-led development exist
A warning about surveillance applications that are possible with current technology
Evidence that AI governance choices have consequences

For readers everywhere:

Neither model treats AI benefits as public goods to be broadly distributed
Concentration of power is a feature, not a bug, of both models
The mechanisms for public input on AI development are weak in both systems

The meaningful question is not whether Chinese or US AI is "better" but what mechanisms exist — or could exist — to ensure AI development serves broad human interests rather than narrow concentrations of power.

Sources

Allen, Gregory. "Understanding China's AI Strategy." Center for a New American Security, 2019. CNAS
BuzzFeed News. "Blanked-Out Spots on China's Maps Reveal the Locations of Uyghur Internment Camps." 2020. BuzzFeed News
Creemers, Rogier. "China's Social Credit System: An Evolving Practice of Control." University of Leiden, 2018. SSRN
CSET (Center for Security and Emerging Technology). "China's AI Workforce." Georgetown University, 2024. CSET
CSET. "Huawei's Progress in Developing AI Processors." Georgetown University, 2023. CSET
DigiChina Project. "Full Translation: China's 'New Generation Artificial Intelligence Development Plan' (2017)." Stanford University, 2017. DigiChina
Feldstein, Steven. "The Global Expansion of AI Surveillance." Carnegie Endowment for International Peace, 2019. Carnegie
Franklin, Ursula. The Real World of Technology. House of Anansi Press, 1989. Publisher
Human Rights Watch. "China's Algorithms of Repression." 2019. HRW
Lee, Kai-Fu. AI Superpowers: China, Silicon Valley, and the New World Order. Houghton Mifflin Harcourt, 2018. Publisher
Miller, Chris. Chip War: The Fight for the World's Most Critical Technology. Scribner, 2022. Publisher
State Council of the People's Republic of China. "New Generation Artificial Intelligence Development Plan." 2017. Translation via DigiChina
Zwetsloot, Remco et al. "China Is Fast Outpacing U.S. STEM PhD Growth." CSET, 2021. CSET

General

China's AI Strategy: State-Directed Development as an Alternative Model