The Chinese Communist Party’s newest AI advance is making repression smarter, cheaper, and more deadly. Even worse, they aim to export it to the world.
March 2025
When news broke in late January 2025 that the Chinese startup DeepSeek had released a new frontier artificial-intelligence (AI) model, it rattled world markets. The open-source model can perform on par with AI giants like ChatGPT, but at a fraction of the cost and computing power. This development put Silicon Valley on its toes. How had DeepSeek managed to achieve the highest AI performance benchmarks with such limited resources? Meta immediately started assembling a team to analyze this remarkable breakthrough. And lawmakers and regulators in Australia, Italy, South Korea, Taiwan, and the United States have initiated measures to block the use of DeepSeek, mostly in government settings.
While the financial, economic, technological, and national-security implications of DeepSeek’s achievement have been widely covered, there has been little discussion of its significance for authoritarian governance. DeepSeek has massive potential to enhance China’s already pervasive surveillance state, and it will bring the Chinese Communist Party (CCP) closer than ever to its goal of possessing an automated, autonomous, and scientific tool for repressing its people.
Since its inception in the early 2000s, the Chinese surveillance state has undergone three evolutions. In the first, which lasted until the early 2010s, the CCP obtained situational awareness — knowledge of its citizens’ locations and behaviors — via intelligent-monitoring technology. In the second evolution, from the mid-2010s till now, AI systems began offering authorities some decision-making support. Today, we are on the cusp of a third transformation that will allow the CCP to use generative AI’s emerging reasoning capabilities to automate surveillance and hone repression.
Situational Awareness
To explain the stages of China’s surveillance state, I use this analogy: CCP governance is like a game of chess — a constant struggle to stay in power. Over the last twenty years, the CCP has increasingly relied on technology to stay ahead in this game. In the first stage, technology helped the party to assess the country (the chessboard): CCTV cameras and data-crunching systems kept law enforcement apprised of citizens’ (the chess pieces’) whereabouts — notifying authorities of the movements of perceived threats to the regime (Uyghurs, Tibetans) as well as perceived allies (mostly Han Chinese). The early surveillance infrastructure was useful because it could tell the CCP what the chessboard looked like at any point in time, and could even predict where a pawn or bishop would move in the near future by calculating its trajectory in a city or by looking at its historical patterns of behavior.
In the early 2010s, Chinese surveillance behemoths such as Hikvision began deploying AI for facial-recognition purposes. Provincial and local governments connected CCTV camera feeds to AI cloud-computing centers, which power citywide surveillance networks. Other “internet-of-things” devices with network connectivity, such as smart thermometers, smart vehicles, and smart watches, added data around the clock. All this information was then fused to personal identifiers, including ID cards and mobile-phone serial numbers, allowing the CCP to anticipate citizens’ behaviors.
With the world’s largest public AI-surveillance networks — “smart cities” — Chinese police started to amass vast amounts of data. But some Chinese experts lamented that smart cities were not actually that smart: They could track and find pedestrians and vehicles but could not offer concrete guidance to authorities — such as providing police officers with different options for handling specific situations.
The Stepping Stone to Autonomous Intelligence
China’s surveillance-industrial complex took a big leap in the mid-2010s. Now, AI-powered surveillance networks could do more than help the CCP to track the whereabouts of citizens (the chess pawns). It could also suggest to the party which moves to make, which figures to use, and what strategies to take.
This transition from tech-assisted to tech-advised surveillance happened over time. But 2016 marked a turning point. That year, the Chinese tech giant Alibaba deployed its City Brain system in Hangzhou, one of the country’s biggest and most advanced metropolises. AI city brains intake, fuse, and process data at an unprecedented speed. Now, if a protest arises in rural Hangzhou, the city brain can immediately alert the local police department, highlight the location of nearby police vehicles, and offer options for responding. If the protest does not reach a certain threshold, however, the city brain might decide to save police resources and not alert forces.
At this point, Chinese tech experts were dreaming of integrating city brains into regional “megalopolis” brains, then a national brain, and one day a world digital brain. In 2021, Guangdong Province took the first step toward a megalopolis brain in the Guangdong–Hong Kong–Macao Greater Bay Area (home to more than 86 million people): a cross-province “one-stop” portal that enables mutual recognition of ID cards, resident household registers, business licenses, and a host of other public services. Among other things, this data integration across provinces facilitates the authorities’ public security work.
In the 1997 book Darwin Among the Machines, the historian George Dyson envisioned the internet as a sentient being that would one day reach artificial general intelligence (AGI) — a vision similar to that of a “world digital brain.” China’s city-brain ambitions have not yet gone that far. But the advent of DeepSeek might propel deliberations within the Chinese smart-city community about whether and how large language models (LLMs) can enable urban environments to reach AGI. Were this to happen, those cities would become vast, digitally connected systems that far surpass human cognitive capabilities.
Inside China, such a network of large-scale AGI systems could autonomously improve repression in real time, rooting out the possibility of civic action in urban metropolises. Outside the country, if cities such as Kuala Lumpur, Malaysia — where China first exported Alibaba’s City Brain system in 2018 — were either run by a Chinese-developed city brain that had reached AGI or plugged into a Chinese city-brain network, they would quietly lose their governance autonomy to these highly complex systems that were devised to achieve CCP urban-governance goals.
Autonomous Intelligence, on the CCP’s Leash
As China’s surveillance state begins its third evolution, the technology is beginning to shift from merely providing decision-making support to actually acting on the CCP’s behalf (that is, playing the game of chess).
DeepSeek is excelling at inference — meaning that its machine-learning model can immediately draw conclusions from brand-new information. It is this technology that would, for example, allow a self-driving car to recognize road signs even on a street it had never traveled before. The company spent remarkably little to develop the model, which is up to fifty times cheaper to operate than its rivals, such as OpenAI’s o1 model. DeepSeek accomplished this using a process called distillation, in which larger models (such as OpenAI’s GPT-4) “teach” smaller models to quickly and efficiently process data.
The advent of DeepSeek has already impelled tech experts in the United States to take similar approaches. Researchers at Stanford University managed to produce a powerful AI system for under US$50, training it on Google’s Gemini 2.0 Flash Thinking Experimental. By driving down the cost of LLMs, including for security purposes, DeepSeek will thus enable the proliferation of advanced AI and accelerate the rollout of Chinese surveillance infrastructure globally.
Looking Toward the Future
The next step in the evolution of China’s surveillance state will be to integrate generative-AI models like DeepSeek into urban surveillance infrastructures. Lenovo, a Hong Kong corporation with headquarters in Beijing, is already rolling out programs that fuse LLMs with public-surveillance systems. In Barcelona, the company is administering its Visual Insights Network for AI (VINA), which allows law enforcement and city-management personnel to search and summarize large amounts of video footage instantaneously.
Although these large AI-powered surveillance systems are currently serving as police tools, in the not-distant future we may see cities deploy AI agents for actual policing. OpenAI’s Operator can already shop online, schedule haircuts, and book vacations. The CCP, with its vast access to the data of China-based companies, could use DeepSeek to enforce laws and intimidate adversaries in myriad ways — for example, deploying AI police agents to cancel a Lunar New Year holiday trip planned by someone required by the state to stay within a geofenced area; or telephoning activists after a protest to warn of the consequences of joining future demonstrations. It could also save police officers’ time. Rather than issuing “invitations to tea” (a euphemism for questioning), AI agents could conduct phone interviews and analyze suspects’ voices and emotional cues for signs of repentance. Police operators would, however, still need to confirm any action taken by AI agents.
The ultimate question for the CCP is how it will prevent its surveillance apparatus from becoming too reliant on autonomous AI. In emergency situations, will police officers check all the incoming information before signing off on an AI police agent’s actions? Or will they simply click “confirm,” since the system always worked in the past? Theoretically, police officers would choose to double-check, vet the system, and maintain control. In practice, however, most would likely opt for convenience and click to confirm without much thought. At the individual level, such overreliance on AI and lack of understanding about how it works might not be an issue. At a systemic level, however, not understanding one’s own governance system will raise serious questions within the CCP.
DeepSeek and similar generative-AI tools make surveillance technology smarter and cheaper. This will likely allow the CCP to stay in power longer, and propel the export of Chinese AI surveillance systems across the world — to the detriment of global freedom.
Valentin Weber is a senior research fellow with the German Council on Foreign Relations. He is the author of the International Forum for Democratic Studies report “Data-Centric Authoritarianism: How China’s Development of Frontier Technologies Could Globalize Repression.” His research covers the intersection of cybersecurity, artificial intelligence, quantum technologies, and technological spheres of influence.
Copyright © 2025 National Endowment for Democracy
Image credit: Nicolas Economou/NurPhoto via Getty Images
|
FURTHER READING |
||
How Autocrats Weaponize AI — And How to Fight Back |
How AI Threatens DemocracyGenerative AI can flood the media, internet, and even personal correspondence, sowing confusion for voters and government officials alike. If we fail to act, mounting mistrust will polarize our societies and tear at our institutions. |
Reimagining Democracy for AIAdvances in AI are rapidly disrupting the foundations of democracy and the international order. We must reinvent our democratic infrastructure to ensure our ability to govern in a dramatically different technological world. |