China’s Surveillance State
For most of the past two decades the Chinese government has been building the largest video surveillance network in the history of computing. The system now extends from the centre of Beijing to the high plateau of Tibet, from the financial district of Shanghai to the oasis towns of Xinjiang, from the apartment blocks of Chongqing to the fishing villages of Fujian. Cameras scan the streets, the alleys, the metro stations, the airports, the schools, the mosques, the monasteries, the malls, the ski slopes, the beaches, the trash cans, and the doorways of private homes. Software identifies the people in those cameras, matches them against databases, and produces alerts for police officers who never see the person being watched. The infrastructure is older than the smartphones most of us carry, the legal framework predates the modern internet, and the political theory that justifies it is older than the People’s Republic itself. Yet the system has been built, piece by piece, with extraordinary speed, and it now operates at a scale that no other country in the world has been able to match.
The purpose of this essay is to describe the system in enough detail to make the technical and political questions legible. We will look at the numbers, the cameras, the software, the databases, the deployment in Xinjiang, the legal framework, the export of the technology to other countries, and the response of the United States and its allies. The picture that emerges is not the picture that the Chinese government presents to its own citizens, which is that the system is a benign set of tools for catching criminals and finding lost children. The picture that emerges is also not the picture that the most alarmist Western reporting presents, which is that every camera in China is connected to a single artificial intelligence that knows everything about everyone. The truth is more interesting, more disturbing, and more important to understand. China has built the most extensive surveillance system ever deployed at scale, the technology is being exported, and the consequences for global norms around privacy and human rights are only beginning to be felt.
The most commonly cited figure for the number of surveillance cameras in China is seven hundred million. The figure is widely repeated and almost certainly understates the total. The four cities at the top of the per-capita ranking, Shanghai, Beijing, Chongqing, and Guangzhou, are reported in independent counts to have more than forty million cameras between them. Shanghai alone has been reported to have more than fifteen million cameras for a population of roughly twenty-five million, a ratio of about one camera for every one and a half residents. Beijing has been reported to have more than eleven million. The most recent official statement from the Ministry of Public Security, issued in early 2026, puts the national total at around one point four billion cameras. The figure is not independently verified, and the methodology by which it was assembled is not public, but the order of magnitude is consistent with the figures produced by Western analysts and with the figures produced by the Chinese surveillance industry itself.
Whichever number is correct, the system is the largest in the world by a wide margin. The United States, the country with the second-largest network, is generally estimated to have around eighty-five million surveillance cameras. The European Union as a whole has fewer cameras than China has in its three largest cities. The gap is not just a function of population. It is a function of policy, of state capacity, of the cost of the technology, and of the willingness of the state to deploy the technology in private spaces as well as public ones. The Chinese system is not, in the main, a system of cameras on street corners. It is a system of cameras inside residential compounds, inside workplaces, inside schools, inside shops, inside mosques, inside monasteries, and inside the rooms of hotels and hostels where the people being watched often do not know that they are being watched.
The cameras are not all connected to a single artificial intelligence, contrary to some Western reporting. They are connected to a large number of different systems, built by different vendors, deployed by different agencies, and integrated at the city, provincial, and national level to varying degrees. The integration is partial, fragmented, and often unreliable. The databases that the cameras query are also fragmented, and a person identified by one system may not be identifiable by another. The system is best understood not as a single artificial intelligence but as a federation of artificial intelligences, loosely coupled, inconsistently maintained, and used in very different ways by different agencies. The federation is what makes the system large. The federation is also what makes the system fragile, because the more loosely coupled the components, the more places there are for things to go wrong.
Skynet and Sharp Eyes
The two programmes that anchor the system are known in English as Skynet and Sharp Eyes. The first, Skynet, is the older of the two. It began as part of the Safe Cities initiative launched by the Ministry of Public Security in 2005, in the years after the SARS epidemic and before the Beijing Olympics. The initiative was originally a response to the perceived disorder of Chinese cities, and it was originally a system of cameras connected by fibre to local police command centres. The system was not designed for facial recognition, because facial recognition software in 2005 was not reliable enough to be useful at scale. The cameras were used for after-the-fact investigation rather than real-time identification. The system was useful for solving crimes, and it was useful for assembling a video record of public life, but it was not yet a system of total surveillance.
The second programme, Sharp Eyes, was launched in 2015, in the years after the Wenzhou train collision of 2011 and in the run-up to the Thirteenth Five Year Plan. The programme was framed in a deliberately Maoist vocabulary, drawing on a phrase attributed to the founder of the People’s Republic about the sharp eyes of the masses. The phrase was a reminder that the system was intended to be participatory, with cameras installed not just by the state but by village committees, by neighbourhood committees, by homeowners’ associations, and by the operators of small businesses. The cameras installed under Sharp Eyes were intended to be connected to local command centres and, through those command centres, to the broader Skynet system. The two programmes are now so closely integrated that they are often referred to together, as the Skynet Sharp Eyes System, and the public documents describing them often treat them as a single network.
The combined Skynet Sharp Eyes network is, by 2026, the largest video surveillance infrastructure ever built. The cameras in the system are owned by a long list of different entities. The Ministry of Public Security owns a fraction. Local police own a fraction. Local governments own a fraction. Homeowners’ associations own a fraction. Schools, hospitals, transport operators, banks, and commercial property owners own the rest. The cameras are connected, in principle, to a national backbone that allows any camera to be queried from any police station in the country. The connectivity is partial. The most recent published estimate is that about seventy per cent of the cameras in the system are reachable from the national backbone at any given time. The other thirty per cent are reachable only from local systems. The two groups together cover, by some estimates, more than ninety-seven per cent of urban public space in China. The rural coverage is lower, but the gap is closing.
The cameras themselves
The two companies that have built most of the cameras in the system are Hikvision and Dahua. Both are based in Hangzhou, in the eastern province of Zhejiang, and both were founded by engineers who had previously worked at state-owned research institutes. Hikvision is now the world’s largest manufacturer of video surveillance equipment by revenue, with a 2025 turnover of about ninety-three billion renminbi, roughly thirteen billion US dollars. Dahua is the second largest, with a similar product mix and a global market share of around ten per cent. Together, the two companies account for more than a third of the global market for professional video surveillance equipment, and they have done so by being both technically capable and politically well connected. Both are state-controlled. Hikvision is majority-owned by China Electronics Technology Group Corporation, a state-owned defence conglomerate. Dahua is closely associated with the Ministry of Public Security.
The cameras the two companies make range from simple dome cameras that cost a few dozen dollars to multi-sensor arrays that cost several thousand. The mid-range products are the most relevant to the surveillance system. A typical mid-range Hikvision or Dahua camera in 2026 is a four-megapixel network camera with infrared illumination, onboard storage, and a small amount of onboard compute. The compute is enough to run a small neural network, and the cameras are designed to be upgraded with more capable inference modules that can run larger neural networks at the edge. The most advanced Hikvision camera, the DeepinView series, integrates a dedicated neural network accelerator that can perform face detection, vehicle detection, and person re-identification in the camera itself, without round-tripping the video to a server. The most advanced Dahua camera, the WizMind series, has similar capabilities. Both products are designed to support the most demanding applications of the Skynet Sharp Eyes system, and both are sold globally.
The companies have been subject to an extraordinary set of sanctions from the United States and its allies. Hikvision and Dahua were both added to the US Department of Commerce Entity List in October 2019, on the grounds that they were implicated in human rights violations against Uyghurs and other Muslim minorities in Xinjiang. The Federal Communications Commission added both companies to its Covered List in 2022, banning new equipment authorisations. The Department of Defense added both to its 1260H list of Chinese military companies in 2023, and the designation has been recertified every year since. The United Kingdom imposed restrictions in 2022 and broadened them in 2023. Canada ordered the wind-down of Hikvision’s Canadian subsidiary in June 2025. India imposed a certification regime in April 2026 that effectively excludes both companies from the Indian market for internet-connected cameras. The restrictions have cost Hikvision and Dahua some of their Western revenue, but they have not stopped the two companies from selling their products almost everywhere else in the world.
The software
The software that turns the cameras into a surveillance system comes from a different set of companies, of which the most important are SenseTime, Megvii, CloudWalk, and YITU. The four companies are collectively known in China as the Four AI Dragons, a term that originated in the early 2010s and reflects the way in which the four companies came to dominate the Chinese computer vision market. All four were founded between 2011 and 2015. All four were built around deep learning research, much of it conducted in collaboration with Chinese universities. All four have been the subject of major investment rounds, and all four have been placed on US sanctions lists.
SenseTime, the largest of the four, was founded in 2014 by Tang Xiao’ou, a computer vision researcher at the Chinese University of Hong Kong. The company specialises in facial recognition, image recognition, object detection, and medical image analysis, and it has increasingly moved into generative AI. By 2025 SenseTime had grown into China’s largest AI software company by revenue, with sales of roughly five billion renminbi, around seven hundred million US dollars, and a market share in Chinese computer vision that has been above twenty per cent for nine consecutive years. SenseTime listed on the Hong Kong Stock Exchange in December 2021, became a partly state-owned enterprise in the years that followed, and has now pivoted substantially into generative AI, with the new business growing at more than seventy per cent year on year in the first half of 2025.
Megvii, the second of the four, was founded in 2011 and is best known for its Face plus plus platform, which is widely used by Chinese developers for facial recognition tasks. The company has been more closely tied to the smart logistics and IoT market than SenseTime, and it has a more diversified product line. Megvii attempted to list on the Hong Kong Stock Exchange in 2019, was caught up in the US sanctions and the post-IPO freeze, and eventually listed on the Shanghai Stock Exchange’s STAR Market in March 2024. The company is a leader in the supply chain and logistics applications of computer vision, and it has a significant business in retail and payment authentication.
CloudWalk, the third of the four, was founded in 2015 in Guangzhou and is best known for its financial services business. The company is reported to supply facial recognition systems to more than four hundred and fifty financial institutions in China, including the six largest state-owned banks, and to have more than seventy per cent penetration of the top tier of the Chinese civil aviation market. CloudWalk has been particularly successful in the export market, with major contracts in Zimbabwe, Belarus, and several other countries. The company listed on the Shanghai Stock Exchange’s STAR Market in May 2022.
YITU, the fourth of the four, was founded in 2012 in Shanghai and is best known for its Dragonfly Eye facial recognition system. Dragonfly Eye is reported to be able to match a face against a database of more than two billion photographs in seconds, drawing on a database of roughly one point eight billion photographs that includes both Chinese nationals and foreign visitors who entered the country through Chinese ports and airports. YITU has had the most troubled financial history of the four companies. Its attempt to list on the Shanghai Stock Exchange was withdrawn in 2021, and the company has been sustained primarily by investment from state-aligned funds. The company’s most notable recent development is a substantial investment from ByteDance-affiliated investors, a sign that the company is now part of a broader Chinese AI strategy rather than a standalone enterprise.
The four companies have all been placed on the US Entity List, on the grounds that they have been involved in the deployment of biometric surveillance technology against the Uyghur minority in Xinjiang. The sanctions have restricted their access to US components and to US customers, but they have not stopped the four companies from growing inside China or from selling to other countries. SenseTime, Megvii, CloudWalk, and YITU remain the backbone of the Chinese facial recognition market, and the technology they sell is increasingly the technology that other countries buy when they decide to build their own surveillance systems.
The facial recognition technology
The facial recognition technology that runs on the cameras and on the central servers is, in its current form, the product of roughly fifteen years of research. The first generation of face recognition systems, in the early 2010s, was based on hand-engineered features and traditional machine learning classifiers. The systems worked reasonably well in controlled conditions, with cooperative subjects and good lighting, but they failed badly in the conditions that the Chinese surveillance system cares most about, with uncooperative subjects, poor lighting, partial occlusion, and large databases. The second generation, which became practical in the mid-2010s, was based on deep convolutional neural networks trained on millions of labelled images. The systems were much more accurate, and the first generation of the SenseTime and Megvii systems, built around these networks, were able to recognise faces in video at near-human accuracy.
The third generation, which is now the standard, is based on more sophisticated deep learning architectures, including the ArcFace loss function, the InsightFace open-source library, and a range of larger and more capable backbones. The third generation systems are able to recognise faces in video at accuracies well above ninety-nine per cent on standard benchmarks, including the Face Recognition Vendor Tests run by the United States National Institute of Standards and Technology. The Chinese systems have consistently been at or near the top of the NIST rankings, often outperforming American, European, and Israeli competitors on the harder parts of the test, which include matching faces captured in unconstrained environments against large databases. The technical achievement is real, and it is the foundation on which the rest of the surveillance system rests.
The third generation systems are not yet perfect. They are less accurate on faces of women, faces of children, and faces of people with darker skin, reflecting the biases of the training data. They are vulnerable to a range of adversarial attacks, including printed glasses, makeup patterns, and projected light patterns. They are vulnerable to liveness attacks, in which a photograph, a video, or a three-dimensional model is presented to the camera in place of a real face. The Chinese companies, like their Western competitors, are investing heavily in liveness detection, in 3D face recognition, and in the use of infrared and multispectral imaging to defeat these attacks. The technology is improving, but it is not yet a solved problem. The systems in production in 2026 are not as reliable as some of the marketing materials suggest, and the false positive and false negative rates are still high enough to produce real harms when the systems are used as the primary input to a policing decision.
The Integrated Command
The facial recognition systems are only the visible part of a much larger data integration system. The Chinese term for the larger system is the Integrated Command, sometimes translated as the Integrated Joint Operations Platform, and it is the system that turns the cameras, the facial recognition systems, the licence plate readers, the WiFi probes, the cell phone trackers, the credit card records, the hotel check-in records, the airline check-in records, the train ticket records, the social media posts, and the neighbourhood committee reports into a single searchable database. The system is not a single piece of software. It is a federation of databases, connected by APIs, indexed by personal identifiers, and accessible from a small number of central consoles in the police stations, the public security bureaus, and the various other agencies that participate in the system.
The Integrated Command system is the part of the Chinese surveillance system that is most difficult for outside observers to evaluate. The system is described in the academic literature, in the investigative reporting of Citizen Lab and IPVM, and in the documents leaked by various Chinese dissidents, but the system itself is not open to inspection. The available evidence suggests that the system is far more capable than the public documents acknowledge, and that it is used in ways that are not described in any official policy document. The most detailed public description of the system is the report by Human Rights Watch on the Integrated Joint Operations Platform in Xinjiang, which describes a system that integrates information about individuals, flags people for further investigation, and generates leads for police action. The system is reported to be in use in at least several hundred counties and cities in China, and it is the technical infrastructure that makes the mass surveillance of the Uyghur population possible.
The database behind the Integrated Command system is built from a large number of different sources. The most important sources are the national identity card system, which contains the name, address, photograph, and fingerprint of every adult citizen in China. The second is the household registration system, which contains the address and family relationships of every resident. The third is the social media monitoring system, which captures posts on WeChat, Weibo, and other Chinese platforms. The fourth is the financial system, which captures transactions and credit relationships. The fifth is the travel system, which captures airline, train, and hotel check-ins. The sixth is the public security system, which captures arrests, court appearances, and other legal events. The seventh is the health system, which has been substantially integrated with the public security system since the COVID-19 pandemic. The eighth is the telecommunications system, which captures the location of every mobile phone in use in China. The integration of these systems is partial, but it is increasing every year.
The grid system and the neighbourhood committees
The Chinese surveillance system is not just a set of cameras and databases. It is also a set of human institutions that operate the cameras, maintain the databases, and act on the alerts the system produces. The most important of these institutions is the grid system, known in Chinese as the ge wei hua guan li system, which divides every city in China into a grid of small geographic units, each containing a few hundred households. Each grid is staffed by a grid manager, a part-time worker who is responsible for knowing the residents of the grid, responding to alerts, and reporting suspicious activity. The grid system is the lowest level of the public security system, and it is the level at which the surveillance system most directly touches the lives of ordinary Chinese citizens.
The grid system is staffed primarily by members of the neighbourhood committees, the residents’ committees, and the village committees that have existed in China since the early years of the People’s Republic. The committees are nominally voluntary, but in practice the members are paid small stipends, and the work is treated as a form of public service that is rewarded with better access to government services. The committees have always been a part of Chinese urban life, but their role has expanded substantially since the early 2010s, and the committees are now expected to play a central role in the surveillance system. The committees operate the Sharp Eyes cameras, manage the grid managers, and act as the first line of response when the system generates an alert.
The combination of cameras, facial recognition, the Integrated Command system, and the grid system is what makes the Chinese surveillance system different from the surveillance systems in other countries. The cameras in the United States, the United Kingdom, and most of Europe are more numerous than they used to be, and the facial recognition systems in use in those countries are increasingly capable. What is different in China is the integration of the cameras with the grid system, the integration of the grid system with the public security system, and the integration of the public security system with the political system. The result is a system in which the state can know, in real time, who is in any public space in the country, and can act on that knowledge through a local infrastructure that has been in place for decades.
Xinjiang
The most extensive deployment of the surveillance system is in the Xinjiang Uyghur Autonomous Region, in the far west of China. The system in Xinjiang was built up over the course of the 2010s, and it has been the subject of more independent reporting and more independent analysis than the system in any other part of China. The reporting and the analysis make clear that the surveillance system in Xinjiang is qualitatively different from the system in the rest of China, in that it has been used as the technical infrastructure for what the United Nations and a number of Western governments have described as a system of mass internment and forced labour.
The Xinjiang surveillance system includes all of the components of the broader Chinese system, and several components that are specific to the region. The Skynet Sharp Eyes network in Xinjiang is, by some measures, denser than the network in any other part of China. The Integrated Command system in Xinjiang is more capable, and the database behind it is more comprehensive. The grid system in Xinjiang has been built up from scratch in many areas, and the grid managers in Xinjiang have been given more extensive training and more extensive responsibilities. The facial recognition systems in Xinjiang have been supplemented by a range of additional biometric collection systems, including mandatory DNA sampling, iris scanning, and voice recording. The systems are integrated with checkpoints, which are positioned on the roads between cities and inside the cities themselves, and which are designed to identify and flag members of the Uyghur population.
The reporting on the Xinjiang system has been extensive. The Associated Press has documented the role of Hikvision and Dahua in the construction of the system. Human Rights Watch has documented the operation of the Integrated Joint Operations Platform. Citizen Lab has documented the use of specific technologies, including the iFlytek voice analysis system. IPVM has documented the specific contracts won by Hikvision, Dahua, and other companies in the region. The Australian Strategic Policy Institute has documented the broader role of the Chinese technology industry in the construction of the system. The New York Times, the Washington Post, the BBC, and many other news organisations have published extensive reporting based on leaked documents, satellite imagery, and on-the-ground investigations. The picture that emerges is consistent across the reporting, and the system has been the subject of a unanimous report by the United Nations Office of the High Commissioner for Human Rights in 2022.
The legal and policy framework
The Chinese surveillance system is built on a legal framework that has been substantially expanded over the past decade. The 2015 National Security Law, the 2016 National Cyber Security Strategy, the 2017 National Intelligence Law, the 2021 Data Security Law, and the 2021 Personal Information Protection Law are the most important pieces of the framework, and they have been supplemented by a large number of sector-specific regulations and local rules. The framework does not, in any meaningful sense, constrain the deployment of the surveillance system. The framework is the legal infrastructure that authorises the deployment of the surveillance system, and it is the legal infrastructure that defines the conditions under which the data collected by the system can be used.
The 2015 National Security Law defines national security broadly, to include political, territorial, military, economic, cultural, scientific, and informational security, and it requires all Chinese citizens and organisations to cooperate with national security work. The 2017 National Intelligence Law requires Chinese organisations to cooperate with intelligence work, and it prohibits the disclosure of the fact that such cooperation is taking place. The 2021 Personal Information Protection Law is the most privacy-protective of the laws, and it imposes a number of requirements on the collection, use, and storage of personal data, but the law contains broad exceptions for national security, public security, and other public interest considerations, and the exceptions swallow most of the protections. The 2025 Public Security Video Surveillance Regulation, which took effect in 2025, is the most specific piece of the framework, and it is the regulation that defines the technical and operational rules for the Skynet Sharp Eyes system.
The legal framework is also the framework that is used to defend the surveillance system in international settings. The Chinese government argues that the system is a legitimate tool for public security, that the data collected is protected by the Personal Information Protection Law, and that the system is necessary to combat terrorism, separatism, and extremism. The argument is not without merit. The system has been used to catch criminals, to find missing children, to respond to disasters, and to manage public health. The argument also does not address the most serious concerns about the system, which are the ways in which it has been used to target the Uyghur population, to suppress political dissent, and to limit the freedom of expression and assembly of all Chinese citizens. The legal framework does not address these concerns because the framework is the same framework that authorises the uses the framework is supposed to constrain.
The export of the technology
The Chinese surveillance system is increasingly being exported. Steven Feldstein’s Global Inventory of AI Surveillance, the most comprehensive public dataset on the subject, documents the use of Chinese AI surveillance technology in sixty-three countries, of which thirty-six are signatories to the Belt and Road Initiative. Huawei is the most prolific exporter, with documented AI surveillance deployments in at least fifty countries. The deployments are concentrated in Africa, the Middle East, and Southeast Asia, but the deployments also include countries in Europe and Latin America. The technology being exported is not, in most cases, the full Skynet Sharp Eyes system, because the system depends on the legal and political infrastructure that exists in China. The technology being exported is the cameras, the facial recognition systems, the Integrated Command platforms, and the integration services that allow the technology to be combined with local databases and local policing practices.
The most important Chinese exporter is Huawei, which has Safe City contracts in more than fifty countries. The largest documented Huawei Safe City contract is the one in Uganda, signed in August 2019 for one hundred and twenty-six million US dollars, which included the installation of eighteen hundred cameras linked to a national police command centre. Similar contracts have been documented in Kenya, Ethiopia, Egypt, Algeria, Angola, Cote d’Ivoire, Zambia, and many other countries. The Wall Street Journal has reported that Huawei technicians have helped local police in at least two of these countries to surveil opposition figures, and the reports have been confirmed by independent investigations. The pattern is consistent across the contracts. The technology is sold as a tool for public safety, the technology is then used as a tool for political control, and the local political outcomes are determined by the local political context.
The second most important Chinese exporter is a combination of Hikvision, Dahua, and the Four AI Dragons. The cameras and the facial recognition systems produced by these companies have been documented in use in Ecuador, Pakistan, Serbia, Uganda, Zimbabwe, the Philippines, and many other countries. The deployments are often less ambitious than the Huawei Safe City contracts, but the deployments are also more numerous. The cumulative effect is that Chinese surveillance technology is now the most widely used surveillance technology in the developing world, and the technology is increasingly being used in ways that reflect the priorities of the Chinese system rather than the priorities of the local legal frameworks.
The chip and the technology supply chain
The Chinese surveillance system depends on a complex global supply chain, and the supply chain has been the subject of a substantial set of restrictions imposed by the United States and its allies. The most important chips in the system are the system-on-chips that integrate the camera, the network interface, the storage, and the inference accelerator. The most important of these chips have been supplied by companies based in the United States, including Intel, Ambarella, and Nvidia. The US restrictions, imposed through the Entity List and through the Foreign Direct Product Rule, have made it increasingly difficult for the Chinese companies to obtain the most advanced chips, and the restrictions have been a major motivator for the Chinese government’s efforts to build a domestic semiconductor industry.
The Chinese response to the restrictions has been to invest heavily in domestic alternatives. The most important domestic supplier of system-on-chips for the surveillance market is HiSilicon, the chip design subsidiary of Huawei. HiSilicon has produced a number of surveillance-focused chips, including the Hi3559 series, which integrates a high-performance image signal processor with a neural network accelerator. Other domestic suppliers include Novatek, which produces image signal processors, and a number of smaller companies that have been acquired or subsidised by the Chinese government. The domestic chips are not yet as capable as the most advanced US chips, but the gap is closing, and the gap is closing faster than many analysts expected.
The other critical components in the surveillance system are the storage systems, the networking equipment, and the software. The storage systems are dominated by a small number of Chinese companies, including Inspur and Sugon, both of which have been placed on the US Entity List. The networking equipment is dominated by Huawei, which produces a full range of switches, routers, and wireless equipment. The software is dominated by the Four AI Dragons and by a number of smaller Chinese companies. The supply chain is now substantially Chinese, and the parts of the supply chain that remain foreign are the parts that are most vulnerable to sanctions.
The Western response
The Western response to the Chinese surveillance system has been slow, fragmented, and substantially inadequate. The United States imposed the first round of sanctions in 2019, with the addition of Hikvision, Dahua, and several other companies to the Entity List. The United States imposed further sanctions in 2020, 2021, 2022, 2023, 2024, and 2025, including the Foreign Direct Product Rule in 2020, the FCC Covered List in 2022, the Department of Defense 1260H list in 2023, and the 2025 FCC proposal to prohibit the continued importation of previously authorised covered equipment. The United Kingdom imposed restrictions in 2022 and broadened them in 2023. Canada ordered the wind-down of Hikvision’s Canadian subsidiary in 2025. India imposed a certification regime in 2026 that effectively excludes the Chinese companies from the Indian market for internet-connected cameras. The European Union has imposed targeted sanctions on individual Chinese officials but has not imposed sanctions on the surveillance companies themselves.
The sanctions have had a real effect on the Chinese companies, particularly on Hikvision and Dahua, but the sanctions have not stopped the Chinese companies from growing. The Chinese surveillance market is large enough to sustain the Chinese companies on its own, and the Chinese companies have made substantial inroads in the developing world, where the sanctions are less of a factor. The most important limitation of the sanctions is that the sanctions do not address the export of the technology. The technology is being exported to countries that are not subject to the sanctions, and the technology is being used in those countries in ways that are not subject to the sanctions. The sanctions are a useful step, but the sanctions are not a solution. The solution would require an international agreement on the export of surveillance technology, and an international agreement is not currently in prospect.
What comes next
The Chinese surveillance system is not finished. The system is still being built, and the system is still being extended. The Ministry of Public Security has stated that the system will cover one hundred per cent of public space in all major Chinese cities by 2028. The facial recognition systems are being upgraded with more capable models, and the systems are being extended to recognise not just faces but also bodies, gaits, voices, and behaviours. The Integrated Command system is being extended to integrate the data with the financial system, the health system, and the education system, and the integration is being done in ways that will make the system more capable and more invasive. The export of the technology is continuing, and the technology is being adapted to the legal and political conditions of the countries that are buying it. The system is not going to stop growing, and the system is not going to become less capable.
The most important question for the rest of the world is what to do about the system. The first answer is that the rest of the world should not buy the technology. The technology is being sold as a tool for public safety, and the technology is being used as a tool for political control. The technology is not a neutral tool. The technology is a tool that has been designed and built to support a particular kind of political system, and the technology is not well suited to the kind of political system that most of the rest of the world claims to want. The second answer is that the rest of the world should build alternatives. The alternatives do not have to be better cameras or better facial recognition systems. The alternatives have to be better institutions, better laws, and better oversight of the technology that already exists. The third answer is that the rest of the world should pay attention. The Chinese surveillance system is the most extensive ever built, and the system is being exported. The rest of the world will not be able to avoid the consequences of the system indefinitely, and the rest of the world should start preparing for the consequences now.
The system is also a reminder of something more fundamental. The technology of surveillance is not, in itself, a new thing. The technology of surveillance has existed in some form since the invention of the camera, and the technology of surveillance has been used by every state in history. What is new is the combination of the technology with the digital infrastructure that makes the technology usable at scale. The combination is not unique to China. The combination is being built in every country in the world, including the United States, the United Kingdom, and the rest of Europe. The difference between China and the rest of the world is not that China has the technology and the rest of the world does not. The difference is that China has decided to use the technology in a particular way, and the rest of the world has not yet decided what to do with the technology. The decision is still pending, and the decision will shape the next century of politics.
