---

Darren Byler is an assistant professor at Simon Fraser University in Vancouver. An anthropologist and author, his research has focused on the dispossession of stateless populations — particularly Uyghur Muslims in Xinjiang — through contemporary capitalism, digital infrastructure and colonialism. He is the author of In The Camps: China’s High Tech Penal Colony (October 2021) and Terror Capitalism: Uyghur Dispossession and Masculinity in a Chinese City (2022).

Q: First I would like to ask about the current state of Xinjiang. There was a recent AP report that described a normalization of fear there. How would you describe what’s happening right now?

A: Things have changed somewhat over the last year. There’s been a new party secretary since January, a guy named Ma Xingrui, and he appears to be drawing down some of the more obvious forms of surveillance — checkpoints and things like that — that were in place. At the same time, though, he and other state leaders are talking about how the counterterrorism campaign will continue, how they’re going to continue to really monitor internet traffic and communication in particular. So that part of the surveillance system is still really in place. And many people that were taken to the camps in 2017 are still missing. They’re still interned somewhere, or some have been formally prosecuted and are now in prisons, often serving prison sentences. Others are in factory settings or industrial parks, where they live and work, and are still quite tightly controlled in a way that’s not that different from a prison-type factory. The system has normalized so that it is now cloaked in a kind of legal frame in terms of people having sentences. But many of the effects are ongoing. And there doesn’t seem to be an end to that in sight.

When was the last time you were in Xinjiang and conducting research from there? 

I was there for the last time in April 2018, doing research. That was already after a lot of people had already been taken, including many of my contacts and people I knew. Part of what I was doing was going to places where people lived and asking the neighbors about what had happened if I could, or just having quick conversations in shops and things like that. Since then, I’ve started to do research across the border in Kazakhstan, with people who have come across the border. So I’m able to corroborate details that I’m reading in internal state documents and other sources through those interviews. But I haven’t been able to travel to Xinjiang recently. Now my information is coming from friends and relatives of contacts, it’s being shared through a kind of chain. And it’s looking at those leaked documents, of which we have hundreds of thousands. 

In 2018, the situation was quite serious in Xinjiang. How were you able to conduct your research safely and interview people, and also keep the people you were speaking with safe?

When I was there in 2018 I didn’t conduct formal interviews. I didn’t have a recorder out or anything like that. I was traveling then as still a Ph.D. student. I hadn’t finished my degree and hadn’t published very much. I don’t think I was really on the radar of the state in a way that I would be now. And so a lot of my research was observational, like going through checkpoints and observing, and because I speak Uyghur and Chinese, Mandarin — or at least have some proficiency in both of them. I’m able to observe and overhear a lot of what people are saying. That’s a lot of what I did. The conversations I had were in the back of unauthorized or informal taxis — they call them ‘black taxis’ in Xinjiang, where they don’t have surveillance equipment because it’s not an official cab or taxi. Also in parks and bazaars when I was buying things, I would strike up conversations in Uyghur very quickly. If you speak Uyghur to Uyghurs, that indicates something about your knowledge of the region. People would often open up pretty quickly and want to share their stories. But of course, these are just really brief conversations, so nothing very in-depth.

A lot of what you do involves looking at a lot of data to understand the level of digital surveillance that has been taking place. Is it a challenge to access that kind of data? How do you find reliable data sources?

The most detailed data that I’ve worked with was obtained by The Intercept. It came from a state contractor called Landasoft, which had contracted with the Urumqi Public Security Bureau. And through means that I really can’t go into detail over, we obtained all these documents and they’re definitely authentic. They’re very detailed. They have the names and ID numbers, phone numbers, geo-locations, and relatives of all these detained people in Urumqi, the capital of the region, and in particular neighborhoods. In those areas we have even more detailed information about exactly how many people are detained, why they were detained, what proportion of the population is detained, what’s happening to their children, what are some of the difficulties that the relatives of detainees are facing, how is the monitoring going in their homes, and are they finding any illegal materials anymore or not — all this really detailed information. It is unwieldy to work with because it’s 52 gigabytes, so it takes computing power. But it’s digitized, so you can run programs with it to find information using keyword searches and things like that. I’m not the only one working on this data. And it will likely take us years to really unpack all of the information and figure out which are the things that really need to be analyzed further and what should be shared with the general public.

Everyone within the Xinjiang region has a digital history that’s being analyzed, and that is pinging their smartphone as they move through space.

So this data came from a private tech company that the government had contracted with?

That’s right. Landasoft had built the mobile policing platform for the Urumqi Public Security Bureau. It’s an app-based interface for the police officers’ smartphones, but also at the fixed checkpoints, it’s used as an interface on desktop computers, and it’s linked up to a region-wide system called the Integrated Joint Operations Platform, which is serviced by another company called the China Electronics Technology Corporation — incidentally, it’s the parent company of Hikvision, one of the largest camera companies in the world, which is also really involved in Xinjiang. Landasoft was contracted with the police to build the system. Using open-source Oracle software, they built this data archive system.

Do we know if this dataset is giving us the full picture of what’s going on in Xinjiang?

It’s certainly a limited data set in the sense that it’s really focused on one city over a particular period of time from 2017 to 2019. It is the largest city in the region. But it’s not the site of the largest number of detentions. The rural areas in southern Xinjiang have much higher rates of detention. And so in that sense, it might be a bit skewed. It’s actually presenting us with a more positive picture, most likely, of how most Uyghurs have experienced the system. It’s still showing us the logic of the system in general, the goals of the system, and its effects. But sure, it is limited. It’s just focused on the 3 million people that live in Urumqi and not the entire 25 million people of Xinjiang.

Is this where that 1.5 million number comes from, regarding the number of people detained?

The number of people detained is an estimate and should be framed as an estimated range of 900,000 to 1.5 million people. And that’s drawing on a bunch of different data sets like the Urumqi data set, but also other data sets from southern Xinjiang.

Your most recent book is entitled Terror Capitalism. Can you describe what exactly ‘terror capitalism’ means and why you characterize what is happening in Xinjiang in this way?

Terror capitalism is a security industrial complex that brings together private contractors — which are technology companies and policing companies — in alliance with the state. I wanted to show that it’s not only an alliance of these different actors, but it’s also part of a global system of capitalism that is linked with some of the logic and some of the technologies of the political system in play in Xinjiang. A lot of the people that are leading these tech companies were trained in the West, in U.S. institutions, or worked in Silicon Valley. And when I talked to them, they say that what they’re doing is really similar to what tech companies in the U.S. are doing to support the U.S. state, the military and the police. They’re saying, “We’re just doing it in China; it’s the patriotic thing to do.” 

The reason why I wanted to call it capitalism and analyze it in that way, is because what’s happening here is there’s capital being accrued, or there’s value being taken through the data that’s being collected from the entire population of Xinjiang. And that data is quite valuable to companies. There’s a direct transfer from state contracting to commercial enterprise in many cases, using the data and the analytical tools they are developing to build commercial products down the line. So it’s a way of rapidly prototyping new technology and then adapting it for other purposes. 

The second aspect of it is that it makes the control of populations much more precise and feasible at scale. Everyone within the Xinjiang region has a digital history that’s being analyzed, and that is pinging their smartphone as they move through space. And in many cases, Uyghurs under the threat of being labeled untrustworthy — the term that was applied to the so-called terrorists sent to camps — are now assigned to work in factories. So there’s a kind of ‘smart factory’ system that’s utilizing this data through checkpoint systems and smartphones. That really holds people in place. It’s aiding in a system of unfree labor. It’s like the Amazon warehouse meets the security industrial complex that is part of a war on terror. And it’s building a new industry of computer vision and data analytics tools in China.

You mentioned that a lot of the technology is being trained on the population in Xinjiang. Are there any technological or AI advancements that we can see today that are a result of how the technology is being used there?

It’s hard to trace algorithmic advancement in a linear fashion. What you can see, if you look at the companies themselves, is that on average within two years of receiving a state contract they develop commercial applications. Because of the contract, they’re given classified access to state data. When it comes to face recognition, it was the base dataset of portraits and iris scans associated with the ID cards of everyone in the Xinjiang region, which was really lucrative and useful. These images were collected as part of a third-generation ID system that utilized biometric data. Everyone was asked to go to the local government office and have their data taken. This often also involves an iris scan, DNA collection, fingerprints, and voice recordings, so they’re getting voice signatures for each person. 25 million people presented their data all around the same time and in a really symmetrical way because it’s standardized. So it’s a really good dataset to work with if you’re a tech company, to develop new tools. And they’re also receiving really large contracts at the level of a city to build a ‘safe city system.’ 

I suppose the largest breakthroughs that have come out of this are around image, face and voice recognition. Companies like iFlytek and Meiya Pico, which is a digital forensics company, are both involved in ‘dataveillance’ tools that scan through people’s phones. And you see scholarship around this time — this is 2015 and 2016 — they’re publishing articles about how technologies can now detect Uyghur text in images, or they’re now able or trying to detect ethnic differences based on facial scans.

You see a story there that’s been told in many, many locations of settler colonialism. A colonial demand for resources and disposing of the people that are in the way of that process.

So part of what I’m saying in the book is that tech companies are actually producing terrorists. That’s their job: to find and manufacture terrorists using data analytics because that’s what the state wants them to do. That’s how you get to a population of 900,000 to 1.5 million terrorists or extremists, which is just unheard of anywhere else in the world.

The use of technology doesn’t seem too different from how it’s used in the U.S., algorithms or technology targeting people of color. 

That’s absolutely true. Some of these tech companies that work in Xinjiang would say explicitly, “We see what Palantir is doing in the U.S., and we need to build tools that are similar to this, but adapt them to ‘Chinese characteristics’ and the data that we have.” The kinds of tools that are developed by state security in the U.S., states around the world are paying attention to that. In order to be a global power, you should develop these tools. In some ways, Xinjiang is a really controlled, preventative policing, domestic war, where there’s very little cost to the people that are perpetrating the war. 

Uyghurs are not a real threat in an organized way at all. But it becomes a testing ground to build these new tools. And the reason that this is happening in the first place is that the Uyghur homeland is the source of 20 percent of Chinese oil and natural gas, and about 85 percent of China’s cotton. The Chinese population that’s been moved there are wanting access to it, and the Uyghurs are seen as threatening that access. So state security is really about securing access to these resources. You see a story there that’s been told in many, many locations of settler colonialism. A colonial demand for resources and disposing of the people that are in the way of that process.

Are there any foreign or U.S. tech companies, or investors, that have effectively aided in the internment of Uyghurs in Xinjiang?

There are different levels of aid or complicity. A lot of the support that companies in China have received is through the training of people; but the training happened before the people started to work as Chinese state contractors, so can you really hold U.S.-based institutions liable for this training? There’s also a lot of use of technology that has multiple uses: such as microprocessing chips from Intel or a supercomputer from IBM. These are really powerful tools. They’re not consumer grade, they’re often sold through commercial service providers or retailers based in China. Companies like Oracle, which is the company that provided software to build the data set I am working on, that Oracle software is open source. Anyone can get access to it. So they would say, [in effect] “How can we police what we make freely available?” For them, it’s about growing their market share and getting more customers. 

…Europe and North America really need to play a larger role in helping shift public opinion in a global sense on this issue, to put concentrated pressure on China. That’s what is really needed to make progress: building more alliances and building an active response.

The critique has to be a little bit more fundamental, about what things should be made open source, should we have data sets of millions of faces made freely available to anybody who wants them? How are we training our computer scientists to think critically about what things like face recognition can do to vulnerable populations? What are the norms of policing? Should we have face recognition to begin with? These are the more fundamental questions. Really what’s happening now is that the technology is developing so quickly that there really isn’t regulation around a lot of this. The companies are figuring things out as they go. And there is an ethos in a lot of tech development around rapid prototyping and just seeing what works, probably a few broken eggs along the way. The problem is groups like the Uyghurs can be harmed by this in deeply damaging ways. And other states can adapt these tools for similar purposes. There’s a lot of harm already and probably more harm to come before things start to get better.

Have you seen any shift in the conversation, or in the level to which these companies are making their technologies available, since more information has emerged about how it’s being used in China?

Yes, I am seeing shifts, though it’s not clear what’s causing them since the companies refuse to take responsibility for their complicity. But the data sets that were used by most of the face recognition companies that worked in Xinjiang, from places like the University of Washington and Duke University and Microsoft and other places, they’ve all been taken down. They’re no longer available open-source. Microsoft Research Asia, which is based in China and is talked about as the cradle of AI development for China, they’ve now said that they’re no longer going to recruit interns or fellows from schools that are affiliated with Chinese state security or with the military. That’s a step in a certain direction. 

Companies like Oracle have also said they’re not the worst offenders. Their response [has effectively been], “Sure, we’ve actively tried to sell our databases to police departments in China, but look at what Microsoft Research Asia has done. Look at all the other ways that tech companies are complicit.” At the same time there’s the other conversation, which is the tech cold war, and how the U.S. has to compete with these Chinese firms. That has accelerating effects back in the U.S. when it comes to developing surveillance technology. And I see that as a race to the bottom, or a race in the wrong direction. We should really be thinking about designing for a better future, not about reviving or building rivalries, building war machines. War inevitably harms those who are least protected.

The U.S. is reportedly considering sanctions against a company you mentioned earlier, Hikvision. There were also the sanctions that were imposed on products from in Xinjiang in December. What do you make of these actions? Are they going to make a significant impact at all, or have they made any impact already?

Many of these companies that I’ve mentioned are on the Entity List from the U.S. Department of Commerce, which means U.S. companies can’t provide products or services to them; but it doesn’t prevent those Chinese companies from continuing to sell products in the U.S. and it doesn’t criminalize actors outside of the U.S. who work with them. My understanding is that if Hikvision is targeted with Magnitsky sanctions and added to the Special Designated Nationals (SDN) list, it would really increase the penalties that people would face for doing business with them. And it would have a pretty major effect on the company, but also on China-U.S. relations. Hikvision is a subsidiary of the China Electronics Technology Group Corporation (CETC), which is one of the largest police contractors in China. And when you see policing systems being built in other places like Ecuador or Uganda, and a few other locations, the export-oriented partner company of CETC — China National Electronics Import & Export Corporation (CEIES) — is often involved, using the same, or related, technology. So Hikvision is probably the largest offender when it comes to human rights violations. And its relationship with CETC helps to establish it as a company that, if there is going to be one that is sanctioned, it should be the one that receives the sanction. 

Could it be more impactful to place sanctions on tech companies in China rather than, say, cotton products or any other particular products coming from the region?

Yes, it probably would be. There’s already a lot of shifting of supply chain manufacturing related to cotton with global brands and their supply chains, with companies like Nike or H&M so far. But it hasn’t really changed the landscape that much in Xinjiang. It’s just that now, Uyghurs are mostly working for companies that are supplying other brands. But it is an important step in reducing complicity in the world system. 

Should countries like the U.S. and other states be going a lot farther in sanctioning or decoupling from China, as we’ve done with smaller countries in the past like South Africa? 

It’s possible. China is a very large country with 1.4 billion people. Xinjiang represents a large landmass, about a sixth of the landmass of China. But it only has a population of 25 million. Most Chinese people see Xinjiang as distant from them, as not something that they feel personally responsible for. So in this sense China is in a different position than South Africa, which was founded as a colony and had an apartheid system for the entire country. 

Full decoupling would have quite negative effects on China. There are hundreds of millions of people that are still in quite severe conditions of poverty in China. And they, of course, would be the most affected by blanket sanctions. 

Regarding sanctions or boycotts related to Xinjiang, really driving home that it is the site of harm is probably what we need more of. Getting the developing world on board, that approach will also be crucial because right now China has quite a large sphere of influence in global bodies like the UN, and in the global South, China’s development is really crucial for a lot of low-income economies and smaller nations. So Europe and North America really need to play a larger role in helping shift public opinion in a global sense on this issue, to put concentrated pressure on China. That’s what is really needed to make progress: building more alliances and building an active response.

---