‘Because There Were Cameras, I Didn’t Ask Any Questions’

Chinese Government Documents Provide New Details on a Small Xinjiang Town’s Extensive System of Surveillance

Sometime in the summer of 2019, Vera Zhou, a young college student from the University of Washington, forgot to pretend that she was from the non-Muslim majority group in China, the Han. At a checkpoint at the mall, she put her ID on the scanner and looked into the camera. Immediately, an alarm sounded and the guards manning the equipment pulled her aside. That was when she remembered that when she ventured outside the jurisdiction of her police precinct she should pretend that she had forgotten her ID and hold her head up high, playing the part of a wealthy, urban Han college student who couldn’t be bothered by mall security and face scans.

In fact, as much as Vera could pass as Han—she liked to wear chunky silver earrings, oversized sunglasses, and dress in black—her ID card said she was Hui, a Chinese Muslim group that makes up around 1 million of the population of the 15 million Muslims who are the majority of the population in the Xinjiang region. Now, a surveillance system connected to local police detected she had ventured out of bounds. As a former detainee in a re-education camp, she was not permitted to travel to other areas of town without explicit permission from both her neighborhood watch unit and the Public Security Bureau. Recounting the ordeal several months later, Vera told me that as the alarm went off, she felt like she couldn’t breathe. She remembered that her father had told her, “If they check your ID, you will be detained again. You are not like a normal person anymore. You are now one of ‘those’ people.”

Vera was living in her hometown of Kuytun (Kuitun) in Ili Prefecture, an area directly north of the Tian Shan mountains that borders Kazakhstan. She had been trapped there since 2017, when—in the middle of her junior year at the University of Washington, where I was an instructor—she had taken a spur-of-the-moment trip back home to see her boyfriend, a former elementary school classmate. Using digital surveillance tools, the Kuytun police had noticed that Vera had used a Virtual Private Network in order to access websites such as her university Gmail account. Given her status as a member of a Muslim minority group, this could be deemed a “sign of religious extremism.”

Police had contacted her boyfriend, who was Han, while the couple were on their way home from a night at the movies, asking him to stop by the nearest police station. Police told him they needed to speak with Vera, so the couple entered the station. There, police explained they would need to take Vera to a larger station for questioning and that her boyfriend could follow the police van in his own car. But once Vera was in a van, safely out of her boyfriend’s sight, a man working for the police pulled her hands behind her back and handcuffed her. “It felt like a horror movie. Like maybe if I said the right things, I would wake up and find that it wasn’t happening. That was when I started wailing, half-screaming, half-crying. The lead police officer told me, ‘It will be better for you if you shut up.’”

Without formal legal proceedings, for the next several months local police held Vera with 11 other Muslim minority women in a second-floor cell in a former police station. In whispered conversations she learned that others in the room had been told they had committed cyber violations, or had engaged in other activities indicative of future criminality according to broad interpretations of China’s counter-terrorism laws that were established in 2016. As Chinese authorities explain in an official document submitted to the United Nations, some people detained in this manner are deemed to have participated in “extremist” activities “not serious enough to constitute a crime.” Vera told me some of the women had used WhatsApp or registered multiple SIM cards to the same ID. Others had WeChat contacts in Kazakhstan or had traveled to Muslim majority countries like Malaysia and posted images of visits to famous mosques and of people dressed in identifiably Islamic clothing. Some had been detained for going to mosques too often, or for possessing retroactively outlawed religious books or other materials such as DVDs with instructions on how to study the Quran.

Vera spent several months in “re-education”: chanting rules, singing “red songs,” writing “thought reports,” and studying elementary Chinese, under the bright lights of the cell and occasionally in a fortified classroom. Then, she was unexpectedly released and placed on a kind of probation. Staff at her neighborhood watch unit told her she was not permitted to leave her small town. Over time, she told me, she came to understand that the “smart city” features of Kuytun meant that at key nodes such as train stations or shopping streets, ever watchful facial recognition cameras could pick her out of crowds and alert the legions of contract workers employed by the local police that she had ventured out of bounds. Fearing that too many of these encounters might result in her being detained again, and aware that many of her friends were afraid to be seen with her, she began to change her behavior. “I just started to stay at home all the time,” she said.

“You are not like a normal person anymore. You are now one of ‘those’ people.”

After several weeks at home, a senior officer in the local police in her neighborhood learned that she had spent time in the United States as a college student. He asked her to begin tutoring his children in English. She said, “I thought about asking him to pay me, but my dad said I needed to do it for free. He also made food for the police officer’s family, to show how eager he was to please them.” The commander never brought up any form of payment. The re-education camp and surveillance system left Vera alone and isolated as an unfree, unpaid educator and nanny, responsible for the education of the police officer’s family. On many occasions, she stood up in front of her neighbors and confessed the errors of her past, how much she had learned by studying the political thought of Xi Jinping, and her gratefulness to the government for giving her a second chance.

In an attempt to prove that she could be trusted to continue her education back in the United States, she tried to follow the rules and always demonstrate a good attitude. Getting caught at the mall checkpoint was not part of this plan—since her neighborhood watch unit was linked to the “smart city” platform, she knew all of her violations would be recorded. Fortunately, the leaders back in her neighborhood watch unit did not see her unauthorized excursion as a sign of deviance and agreed to let her go with a warning. Eventually, after six months under neighborhood arrest and after signing numerous documents guaranteeing that she would not talk about what she had experienced, Vera was allowed to reclaim her passport.

When Vera emerged from the escalator into the baggage claim area of Seattle-Tacoma International Airport on September 18, 2019, she smiled weakly at me and the small group that had gathered to welcome her home. She looked exhausted. Her life had been forever changed by her time in the system of confinement and surveillance that had overwhelmed her in Kuytun. And all along, she had been living in what China’s government and technology industry calls a “Safe City.”

In combing through hundreds of government documents, I have found that dozens of county-level or larger administrative divisions in Xinjiang have developed “Safe City” surveillance systems. This network of Safe Cities is a particular instantiation of nationwide Skynet, Smart City, and Sharp Eyes systems. Since 2017, when the re-education campaign intensified, these programs have enveloped the entire region—pulling Xinjiang’s surveillance systems into alignment and, in some ways, exceeding the capacities of areas in Eastern China. Stories like Vera’s show how these surveillance systems can control the lives of those they target, but the inner workings of such systems have remained something of a puzzle.

* * *

Newly uncovered procurement notices for the small town of Shawan and the surrounding county, 64 kilometers away from Vera’s hometown of Kuytun, now provide a more fine-grained view of how a Safe City surveillance system in Xinjiang can function. As outlined in nearly 400 pages of a 2017 feasibility study and two legal contracts, the system Shawan’s officials hoped to purchase would be supported by the Face++ algorithm designed by the computer vision company Megvii. Megvii denies it did business in Xinjiang apart from selling facial recognition registration systems that tie hotels to Public Security Bureaus. But an investigation published earlier this month by surveillance industry publication IPVM revealed that Megvii collaborated with Huawei to develop a “Uighur alarm” which would automate the detection of Uighur faces in video monitoring.

The re-education camp and surveillance system left Vera alone and isolated as an unfree, unpaid educator and nanny, responsible for the education of the police officer’s family.

The Shawan Face++ system would be designed to assess object information such as license plates, but also to hone in on “human faces, physical features, accessories, and so on,” the feasibility study explained. It would track those identifiers while gathering other social data such as “communication behavior, accommodation behavior, migration behavior, financial behavior, consumer behavior, driving behavior, and administrative violations.” The accuracy of the system would depend on a base set of images associated with state-issued IDs and comparison technology used to conduct image analysis of other captured images—a system similar to Clearview AI software used by law enforcement agencies in the United States. It would also use analysis of real-time video—something similar to a London police pilot project using NEC-supported video face surveillance. While there are some similarities between American and British policing systems, which also disproportionately harm ethno-racial minorities according to a study by the ACLU, and those in China, the high degree of surveillance density in Shawan, as well as the capacity of its detention facilities, show that there are remarkable differences as well.

The Shawan “Safe City” procurement documents are the first to show in minute detail the design parameters and feasibility of a county-level surveillance system in Xinjiang. In the context of Northern Xinjiang, Shawan is an unremarkable, aging, Han-majority town. According to the Xinjiang Statistical Yearbook for 2018, the total population of the county is around 200,000, with around 67 percent of the population between the ages of 18 and 60 years old. Almost 80 percent of the people in the county are employed by the Bingtuan, a paramilitary farming colony that was recently sanctioned by the U.S. Government for its involvement in Xinjiang’s detention camps and forced labor. Many residents of Shawan are involved in cotton and tomato cultivation—two of the primary crops of the region. The ethnic minority population of the county is around 72,000, but only 9,500 of them are Uighurs. Most of the others are Kazakh and Hui. If demographic trends hold across the entire population, only around 48,000 of these Muslims are between the ages of 18 and 60.

Given that the town appears relatively unremarkable—it has no history of violence, it is simply a typical Bingtuan town—the sophistication of the proposed surveillance system is surprising and suggests that even small Han-majority towns are highly invested in the detention and re-education campaign. Dozens of state media reports and a series of procurement notices from Shawan in 2017 and 2018 make clear that the Shawan Public Security Bureau bought and operationalized significant portions of the surveillance system described in the feasibility study.

It is also clear that many Muslims in Shawan have been detained. While researching this story in August 2020, I came across a brand-new detention facility, built on the north side of town in 2019. Raphael Sperry, an architect and the secretary of the U.S.-based organization Architects/Designers/Planners for Social Responsibility, estimates that it has the capacity to hold between 7,000 and 20,000 people, roughly 15 to 40 percent of all Muslim adults in the county. The new facility’s construction flies in the face of the Chinese government’s public line on the detention of Uighurs. In early 2019, the governor of Xinjiang, Shohrat Zakir, said everyone held in the camps had “graduated.” The dramatic expansion of the detention facility in Shawan, after this announcement was made, belies this claim.

Top image: Google, © Maxar Technologies; Center image: Google, © CNES/Airbus; Bottom image: Google, © Maxar Technologies

Images of what is likely a detention facility being built in Shawan county, accessed on Google Earth December 19, 2020.

Since 2017, Xinjiang has become a limit case for even Chinese surveillance systems. In the region, the networks of cameras are denser than systems in other parts of China and are supported by more than 7,700 surveillance hubs (dubbed “People’s Convenience Police Stations”), thousands of face-scan and phone-scan checkpoints at jurisdictional boundaries, and nearly every resident of the region has submitted their biometric data to the authorities in a comprehensive “public health” initiative. Because of the fidelity and scale of iris scans and facial image portraits from Xinjiang IDs, which form the baseline of the face recognition systems, the tools Xinjiang contractors employ are more fine-tuned and invasive than their counterparts elsewhere in the world. The Megvii algorithm, and the algorithms of its principal rival companies in China—YITU and Sensetime—run “1:N,” or one-to-many, comparisons to hundreds of millions of images at an extremely fast speed. Moreover, when partnered with Xinjiang’s Public Security Bureaus, these algorithms have access to many high-definition images of the faces of each resident of Xinjiang.

In Shawan, the Megvii system would be designed to match the faces, register, and record alarms for up to 300,000 targeted people in .8 seconds, and, in two-tenths of a second longer, 500,000 people. Matches within such systems are always a probability. But technicians in Shawan would be able to “set the comparison threshold manually” to finetune the targets of the surveillance algorithm. And the longer the system would run, the more fine-tuned it would become. Over time, the feasibility study explains, it would automatically “supplement information for portraits in the base library” with views of residents throughout their daily life, and, by collecting on-line behavioral data, “enhance person-specific information.”

According to an additional procurement notice, in September 2017, the Shawan Public Security Bureau began to buy biometric data collection equipment which would “collect and identify basic personal information and social network data” from populations of “registered and unregistered people,” those who had criminal records, the migrant “floating population,” people from other countries, and “two types of prisoners”—a phrase that appears to refer to prisoners in the reform through labor and “re-education” through labor systems (parts of which were officially abandoned in 2013). Another notice explains explicitly that the Shawan Public Security Bureau would use the system to track “detainees and people undergoing education transformation.” Then, on December 11, 2017, in what appeared to be a response to the feasibility study, Shawan issued another major procurement notice announcing the winning bid of the Shawan Safe City project. The notice outlined how the Shawan Public Security Bureau would pay 105 million renminbi to Shenzhen Jiaxinjie Technology to construct the Safe City system. The first year of the project, Jiaxinjie would complete a “front-end construction, transmission network construction, platform construction, command center and sub-command center construction, and monitoring and lighting system construction.” Over the following 10 years, the Shawan Public Security Bureau would pay the Shenzhen company 17.7 million renminbi per year to maintain the system.

Over 2018, the Shawan Public Security Bureau issued a series of additional procurement notices. In another notice posted in January 2018, the Shawan Public Security Bureau declared its intent to purchase 82 sets of iris-scanning equipment from Beijing Wanlihong Technology for 1.4 million renminbi to use in building a new ID dataset. In yet another procurement notice, the Shawan Public Security Bureau declared its intent to purchase 52 sets of “portrait collection devices” and 32 sets of fingerprint collection machines for 1.2 million renminbi from the Third Research Institute of the Ministry of Public Security.

* * *

According to much of the framing of state media reports, the design and implementation of the Safe City project in Shawan created greater governance efficiency. In 2018, the Shawan Public Security Bureau administrative services system—responsible for household registration and residential identification documents—transitioned to a smartphone-driven digital model. A wide array of smartphone-based facial recognition technologies enabled this administrative system to collect individuals’ data, a public service announcement explained. The implementation of smartphone-based systems started with a face-scan-centered state-issued photo ID. By mid-2018, applications to replace lost IDs could be submitted online using a smartphone camera to scan the applicant’s face. In 2020, the Shawan County Human Resources and Social Security Bureau announced that to access social security benefits, residents of Shawan were required to use a new face-scan app.

In order to pass through neighborhood-entrance face-scan checkpoints, residents needed to install another face-scan app on their phones, a government announcement explained. Even agricultural work brigades installed face-scan systems to monitor villager work efficiency. Students and parents installed yet another face-scan app that allowed them to enter school grounds and track student behavior. If the feasibility report was implemented as planned, much, if not all, of this face-scan and phone data from housing areas, work units, and schools would be integrated into the broader three-tiered Safe City system through the “societal resource integration platform” and funneled through 14 new neighborhood watch unit “sub-command centers.”

As outlined in a recent ChinaFile article by Jessica Batke and Mareike Ohlberg, the proposed Shawan system would use 4,791 networked HD cameras, 70 of which were to be facial recognition units.

[It] would be positioned in crowded places with clear entrances and exits, including mosques, with others to be installed in train stations and bus stations. On the back end: a set of interlocking platforms would span three administrative levels (the village/township level, the county level, and the prefectural level) and three network layers (the public Internet, a private video network, and the Public Security Bureau’s own intranet). Critically, the system would allow for information to flow from private cameras to the police via a “societal resource integration platform” that drew from surveillance in “hotels, Internet cafes, gas stations, schools, hospital monitoring [sic], bicycle rental points, and shops along the street, etc.”

Some locals seemed to welcome the new technology. For example, in 2020, a real estate company in Shawan advertised face-scan technology as a point of convenience and security. The real estate company also presented as a selling point that all urban housing in Xinjiang is monitored by neighborhood watch units under a grid management system.



State of Surveillance

Jessica Batke & Mareike Ohlberg
Across China, in its most crowded cities and tiniest hamlets, government officials are on an unprecedented surveillance shopping spree. The coordination of the resulting millions of cameras and other snooping technology spread across the country...

But newly built forms of surveillance also seemed to inconvenience some residents. For instance, as new systems were placed in motion in 2017, the Shawan authorities began to conduct “night patrols” that required drivers to stop and be inspected. This prompted a resident of Shawan using the surname Hu to complain to his friends on WeChat. According to a state media source called Peaceful Shawan, the Shawan “Internet Security Police”—a division of the Public Security Bureau stationed in command centers in the Safe City system—discovered these private posts, detained him for 15 days, and publicly shamed him. In 2019, the Shawan police also noted the need to “debug” the system so that all legal residents could be automatically identified at checkpoints, assuring that government and technology company personnel would not need to run manual checks.

* * *

The Safe City system needs human monitoring to operationalize it, and, at times, this human intervention creates friction. As in all complex technology systems, human technicians are required to finetune the data and “debug” the system. Technology studies scholar Lilly Irani describes these workers as “data janitors.” In this context, the data janitors of the Safe City system in Shawan and throughout the region were 90,000 police contractors or assistants (xiejing, 协警) hired at the beginning of 2017. According to job listings, most of these recruits would not receive formal training in police academies as Public Security Bureau employees do. Most would not be authorized to carry lethal weapons. In other places in China, they would simply be referred to as “security guards” (bao’an, 保安) or “urban management officers” (chengguan, 城管), but in this context they had power over Muslim life from their positions in People’s Convenience Police Stations. According to research I conducted in Xinjiang in 2018 as well as dozens of online advertisements, local Public Security Bureaus hired many of them from Muslim minority populations. The basic qualification for the job was having a “trustworthy” family background, actively opposing “ethnic division and illegal religious activities,” and having a basic working knowledge of Chinese. The ads actively recruited the spouses and children of “Xinjiang Aid” (Yuan Jiang, 援疆) personnel. These mostly Han citizens were brought to Xinjiang as part of a “paired assistance program” that linked wealthy cities and provinces in eastern China with particular counties and prefectures in Xinjiang. This “poverty alleviation” scheme brought industries, educators, and labor transfer programs to Xinjiang, but as the scholar of Chinese politics Jennifer Pan points out, it also brought new forms of surveillance and coercion. Shawan’s “sister city” was Anshan, a city of around 3.5 million people in Liaoning province, in northeast China.

As one police chief in Urumqi put it, in general, the job of the new contractors is to create a 24-hour “seamless” security environment. According to Baimurat, a former police contractor who spoke about the system in an interview posted online, assistants conduct spot checks centered on actively profiling passersby, stopping young Turkic people and demanding that they provide their state-issued IDs and open their phones for automated inspection via spyware apps and external scanning devices. Policing contractors monitor face-scanning machines and metal detectors at fixed checkpoints. All of these activities assure that information forcibly collected from Uighur and Kazakh residents continues to feed the dataset of the system, making “extremism assessments” conducted by neighborhood watch units more and more precise. As anthropologists such as Joanne Smith Finley have shown, Muslims determined to be “untrustworthy” through data checks can be sent to detention centers where they are interrogated, asked to confess their violations and name others who were also “untrustworthy.” In this manner, the parameters of the technopolitical system determine which individuals are slotted for what was referred to in Shawan as a “centralized closed education training” internment camp.

* * *

On the edge of Bingtuan farming colonies, in another small town called Qitai, 370 kilometers directly east of Shawan, Baimurat was one of the police contractors who conducted these checks. He was in one of the first groups of contractors hired from across the region in late 2016. In an extensive, hour-long Kazakh language interview posted online after he fled across the border to Kazakhstan in early 2019, Baimurat provides fine-grained detail on how surveillance systems have overwhelmed small Xinjiang towns and are linked to the camp system. He is one of only a handful of former security workers in the Xinjiang surveillance system who has spoken to researchers and journalists.

“We handcuffed and shackled them and then we gave blankets to them whether they could hold them or not, and we told them to get on the bus. I had to handcuff one person that I had a feeling I had seen before. Then I realized he had worked as a police contractor as well. I had seen him before while I was working. I didn’t remember his name, but I knew him. I really wanted to ask what happened to him, but because there were cameras I didn’t ask any questions.”

In the interview, he says that because he was a college graduate he was “considered very well qualified.” As a result, he received the highest-level salary available to the contractors, around 6,000 renminbi per month, which is far above the minimum wage of around 1,800 renminbi. Others in his cohort who were considered less qualified because of their educational background, he says in the interview, were paid closer to 2,500 renminbi. Through research conducted in Kazakhstan in early 2020, I confirmed details of Baimurat’s story through an intermediary. Alongside past reporting done by The New York Times, I pieced together that, like many highly educated Muslims in northwestern China, Baimurat had struggled to find work for which he was qualified in the past, so taking the job was a choice he felt he could not refuse. Not only would he be able to provide for his family, but he would also be able to protect them from the re-education system. “We were given uniforms,” he says in the interview. “Then we started doing different kinds of training. It was really strict, as if we were planning for a war.”

In late 2016, Qitai authorities started building People’s Convenience Police Stations, a type of surveillance hub erected every several hundred meters in Muslim majority areas as part of the Safe City grid. Authorities divided the contractors up and stationed them at one of the 89 stations that were built in Qitai County.

According to local media reports and hiring ads, in Shawan a similar process played out in 42 stations erected in 2017. The Public Security Bureau of Shawan recruited at least 490 auxiliary officers. According to one ad, 316 would be stationed within the town of Shawan along with 48 ethnic minority recruits, and the remainder would be stationed in rural stations. There is some indication that the force continued to expand. In 2018, Shawan authorities circulated advertisements first for 100 more officers in the spring and 200 more in late summer. In 2019, Shawan recruited another 200, and through a series of additional procurement notices announced that it had awarded bids for close to 7 million renminbi of equipment and other supplies related to the People’s Convenience Police Stations. The Shawan feasibility report from back in 2017 called for 77 stations.

Back in the Qitai stations, “initially, we sat facing the TV monitors, and you could see the places where the cameras were pointed,” Baimurat recalled in the online interview. “We had to sit there monitoring them all the time. If we failed to notice an alert, or stopped looking, we would be punished.”

Over time, higher-ups in the Public Security Bureau began to assign different types of surveillance work to Baimurat and the other police contractors who were assigned to the People’s Convenience Police Stations. First, superiors sorted the contractors based on their Chinese language ability, demonstration of loyalty, and how well they understood the concept of “extremism” and its elastic definition in the context of the re-education system.

“They made us do other exercises like reciting rules about participating in the camp system,” Baimurat said, in the interview. “We had to recite things related to law. There were quotes from Xi Jinping on the walls of the station. We had to learn these by heart. We were not allowed to go out on patrol until we successfully recited the quotes from Xi Jinping.”

Then, around the middle of 2017, the police contractors were tasked with actively finetuning the programming of the Safe City system using digital forensics tools. Plugged into smartphones through a USB cable, these devices scanned the phones’ files searching for up to 53,000 markers of Islamic or political activity.

After I had been working there for six months, they handed out devices to check pedestrians and car drivers. When we scanned their ID card and phone with it, we got information about whether or not the person had worn a veil, had installed WhatsApp, had traveled to Kazakhstan. All sorts of things like that.

They began to perform night checks, like the ones that had upset the person surnamed Hu in Shawan. “We could stop every car on the street and check them. When we stopped them, we asked the people inside to show their phones and ID cards. If there was something suspicious like I mentioned before, we needed to inform the leaders.”

Around this time, Baimurat learned that although he was a contingent worker, hired on a contract basis, he was not free to quit. “If we were tired and wanted to quit, they would tell us if you are exhausted you can take a rest, but then you must come back. If you quit the job, then you will end up in the ‘re-education camps’ too.”

Initially, Baimurat and his coworkers felt that despite the long hours and the confrontational positions they were placed in, being a police contractor “was a good job” with a steady paycheck and protection from police harassment. They saw themselves as on the side of the “good guys.” This began to change around the time that they received the smartphone-scanning equipment. He said:

I learned then that they had sent the children from the Kazakh Number 3 Middle School in the county seat to the Han school. They built an iron gate, high electric fence, and four watchtowers around the Kazakh school. If we found anyone suspicious through the ID checks, they would send them to the Kazakh school. They had suddenly turned it into a prison. They forced all of the people who had been visiting mosques, praying, or wearing headscarves to go to that school.

Initially, it seemed to him that it was just people who had been actively religious who were sent to the new “prison” school. It was close to six months before he fully realized the implications of the “rounding up those that need to be rounded up” policy that the Xi administration had implemented through a speech given by Xinjiang Party Secretary Chen Quanguo in February 2017.

While I was working one day, we had a meeting. It was in early 2018. In the meeting, we were told we had to transfer some detainees from the jail to the school. We had so many manacles. When we got there, we saw that they had caught around 600 people. There were rooms inside the building that were like cells. I saw very young women, very old women, and men with beards [over the age of 55] among the detainees. They were mainly minorities, the majority were Uighurs, then a few Kazakhs and some Hui people. I don’t think there were Han people. Maybe one or two, but not more than that. We handcuffed and shackled them and then we gave blankets to them whether they could hold them or not, and we told them to get on the bus. I had to handcuff one person that I had a feeling I had seen before. Then I realized he had worked as a police contractor as well. I had seen him before while I was working. I didn’t remember his name, but I knew him. I really wanted to ask what happened to him, but because there were cameras I didn’t ask any questions. I thought maybe I could ask later. But I never found a chance.

In the interview, Baimurat spoke in a quiet voice without much expression on his face. This began to change as he discussed this moment of encounter. He held his hands out in front of him, showing the way the detainees were shackled and how they threw the blankets on their bound hands.

Continuing, he explained that sometime later, when he felt it was safe, he asked another Kazakh police contractor about the man he had recognized among the detainees. His coworker told him, “he came from a village and didn’t understand how the CCTV cameras worked. While he was working in the prison, he saw a paper on the floor which said ‘get me out of here.’ He didn’t report it, but the camera saw it so he was taken ‘to study.’” Hearing this, Baimurat said that for the first time he fully realized that any Kazakh or Uighur could be sent to the camp. No one was safe, no matter how hard they tried to work within the system. He said:

I felt very bad about being part of the system. There were so many people who made very tiny mistakes and ended up there. As police, we had tasks we were forced to do. Some days the leaders said do this, other days they said do that. Each day, we had to do what they said.

Over time, the pressure wore on Baimurat and his wife. He said, “We couldn’t sleep. We were crying all the time, my wife and I. But we didn’t show other people that we were crying, because they might think we were dangerous and might inform on us.”

* * *

Back in Shawan, as the Safe City project was implemented, a similar process appeared to unfold. On the weekend of April 7, 2017, the leaders of Shawan county attended a meeting where Chen Quanguo declared a new beginning to the ongoing “de-extremification” campaign. The cadres said that they would renew their resolve to “resolutely oppose the infiltration of religious extremist ideas.” By August 2017, the county jail had been expanded and officially turned into a “concentration re-education center.” In early 2018, two of the mosques in the town were destroyed, and a third had its Islamic architectural features removed.

Over the years that followed, the Xinjiang Victims Database, an international organization, has documented 114 cases of people detained in Shawan. In the majority of cases, the relatives of Shawan detainees said they had no knowledge of why their relatives were detained. In some cases, detainees or their family members said they were detained for violations that did not rise to the level of criminality and were related to online activity or travel to Kazakhstan to visit their relatives. The most frequently known reason for detention, however, was “praying” and “mosque visitation”—the site where the majority of facial recognition cameras were slated to be installed in Shawan’s Safe City project.

Courtesy of Orazhan Qunapia

Mahmutjan Abla and Nurbai Qunapia in an undated wedding photo, provided to the author by Nurbai’s sister, Orazhan Qunapia.

In Shawan, it appears that first they came for the Uighurs. One of the first people detained was a Uighur baker named Mahmutjan Abla. He was taken in the fall of 2017. The police told his wife it was because of his devotion to Islam. Then in March 2018, they came for his wife, a Kazakh woman named Nurbai Qunapia. It was around this time that the Shawan Public Security Bureau began to take most of the other documented Kazakh detainees.

When I interviewed Nurbai’s sister, Orazhan Qunapia, in January 2020 about what had happened to her family, she said her relatives had been taken for “no real reason.” They practiced Islam and prayed five times per day, but so did most of the 48,000 adult Turkic Muslims living in Shawan county. What may have set them apart was that Mahmutjan was Uighur and had gone to visit Orazhan’s family in Kazakhstan in 2014. “Actually, they both had gotten their passports, but Nurbai never had a chance to come,” Orazhan said.

The couple had met and fallen in love when they were both working as traders in Urumqi. Since they were from different ethnic groups, their marriage was a bit unusual. They had a “love marriage” that was not arranged by their families. “Nurbai was independent. She also had a small business when they met. She sold clothes in the market to make money for herself.”

Even more remarkably, given Uighur patriarchal traditions, Mahmutjan decided to move to Han- and Kazakh-dominated Shawan, rather than back to his home village in Khotan prefecture. He even had his household registration changed to Shawan. Thinking about Mahmutjan’s visit to Kazakhstan in 2014, Orzhan remembered. “He was funny. Since he was a baker, when he came here he really just wanted to see how people bake bread in Kazakhstan. He loved to learn new things.”

For the first time, he fully realized that any Kazakh or Uighur could be sent to the camp. No one was safe, no matter how hard they tried to work within the system.

After Nurbai was taken in April 2018, Orzhan’s communications with the couple went dark. Then, without warning, Nurbai was released in December 2018. “We talked via WeChat in December. We talked about her release. She said that Mahmutjan was still in the Shawan camp. I congratulated her on surviving. She said, ‘but my health is not good.’ We had been careful not to talk about any political or sensitive subjects, so when she said this I knew that she must have been deeply hurt by her time in the camp.”

Within several weeks, Nurbai disappeared again. Orzhan suspects that, like others in Shawan, she was forced to work in a factory, perhaps at the Shawan Textile and Garment Industrial Park that manufacturers from Anshan city set up as part of their “Xinjiang Aid” project. In other parts of Xinjiang and across the country, these “aid” programs have been shown to be part of Muslim forced labor schemes. In 2019, the broader “poverty alleviation” scheme in Shawan resulted in county authorities assigning 15,600 “surplus laborers” from Shawan to jobs in places like the industrial park. As I have shown in other research, in the context of Xinjiang’s camp and surveillance systems, poverty alleviation refers to moving Muslims from farming and self-employment into wage labor employment, particularly in factory work. There are three main tracks used to accomplish this transformation. The first is through the camp system itself, the second is through local authorities assigning underemployed Muslims to work in factories, and the third is by building “satellite factories” in rural areas. In Shawan, it appears as though the first two tracks were the dominant ways that Muslims were put to work.

What has troubled Orzhan the most is that Nurbai and Mahmutjan have three children. “Since December 2018, we have not heard any news about them. No one knows where the children are. They have two daughters and one son. The oldest is Dilhumar. She is 22. The second is 18 years old, named Dilmira, and the youngest, Enser, is seven years old. The oldest has graduated from school. The second is in school. We don’t know where the youngest is. We don’t know how he is surviving. It has become almost impossible to find out any information. We can’t call our relatives.” Orzhan worries that Enser, the little boy, has been taken to an orphanage like other children have been in Shawan.

Orzhan also worries that her brother-in-law may have been given a prison sentence like so many other former detainees.

Google, © Maxar Technologies, CNES/Airbus

An image of Shawan’s county seat and surrounding areas, accessed on Google Earth December 17, 2020.

* * *

In the feasibility report describing the parameters of the Shawan Safe City Project, the assessors argue that the project will “greatly improve the local government’s ability to quickly respond to major events and ensure economic construction.” At the same time, they suggest it will also foster private business investment and improve social security by protecting private property. “The indirect economic benefits are immeasurable,” they write.

In April 2017, two months after Party Secretary Chen Quanguo declared the new “round up those who should be rounded up” campaign and the Shawan procurement notice was posted, the Party Secretary of Tacheng prefecture, a man named Xue Bing, visited Shawan and toured the different facilities involved in the re-education campaign. According to a report from Shawan News, a local state-affiliated weekly news service, he visited the Public Security Bureau command center to inspect the “implementation of stability maintenance work.” In his remarks to town officials and state workers, he discussed strengthening “education and management services” for mosques—two of which would be destroyed the following year. He described the Shawan Textile and Garment Industrial Park as a key area for future ethnic minority employment. Meetings conducted at the same time among Shawan Party leaders emphasized “study transfer” programs and that cadres tasked with village surveillance should fully embrace their role in waving the “assessment baton . . . and enforce strict discipline in the villages.”

Weekly updates in Shawan News between April and December 2017 suggest surveillance and political campaigns suffused Shawan. Announcements and advertising from work units and housing complexes make clear that nearly every aspect of the lives of Shawan’s residents—from the apps on their phones, to the ways they enter and leave their homes—depends on a set of digital codes linked to images of their faces. If the plan laid out in the feasibility study was implemented, these codes would function as flexible systems of monitoring, enclosure, and blockage—the checkpoints could be turned on and off, the sensitivity of watch lists could be manually adjusted. The data collected would then circulate on the regional intranet, among police contractors and Public Security Bureau employees who are themselves surveilled by the system.

As my conversations with Vera make clear, because of this segmentation, the algorithms of face-codes deny the people they target from knowing how and why they are being surveilled. Those questions cannot be asked. What is known is that they make people discipline themselves by monitoring deviance in their daily routines. They reroute people, separating sisters from each other and parents from their children. As Baimurat noted, they produce all of the feelings of walls and watchtowers, but move beyond the sense of being blocked and watched to the ability of Muslim residents to communicate and influence others. Ethnic minority farmers and self-employed business owners like Nurbai and Mahmutjan have been removed or walled off from the protected population by this system.

At the same time, those that are protected by these systems are, despite minor inconveniences, generally empowered by them. In Shawan, real estate appears to be more valuable because of them. According to the feasibility study, street lights would have been installed throughout the county, making it harder for people to escape notice when dumping their trash.

Taken as a whole, the interlinked surveillance and camp system in Shawan appears to center on assuring the movement of “beneficial” goods, services, and biometric data, while channeling or stopping the movement of objects, bodies, and data which could potentially disrupt this circulation. From the perspective of its designers and purchasers, it would increase the circulation of what the state and its market economy deem “good” and cut off the movement of anything and anyone they reject. The shaping power of this technology would regulate the population to produce lower rates of fear, higher rates of economic growth, and greater power for the right kind of people.

Baimurat recalled that a fellow Kazakh contractor, a 24-year-old Tianjin University graduate with a fancy car, made the mistake of giving a Uighur detainee a cigarette during his midnight shift at the camp. For young Kazakh and Uighur men, sharing a cigarette is a basic sign of respect. “He thought it wouldn’t be a big deal, so he did it. But he forgot about the cameras. Just as the man finished smoking it, the higher level police came and took my coworker away. We never saw him again.” In the control society of a Safe City, life is made predictable by maintaining relations of power at a technical remove. It does something more than this, too. When the technology begins to think for people, it starts to strip away their basic humanity. When space for thinking is lost in the black box of a complex technological system, it becomes banal and inhumane, producing profound capacities for cruelty.

Jessica Batke provided research for this article.

Correction Note: A previous version of this article misstated the location and number of testimonies documented by The Xinjiang Victims Database. The organization is internationally based and has collected 114 testimonies.