The Karakax List: Dissecting the Anatomy of Beijing’s Internment Drive in Xinjiang

Journal of Political Risk, Vol. 8, No. 2, February 2020

Figure 1. The first (redacted) page of the 137-page PDF. Source: Uyghur Human Rights Project.

Dr. Adrian Zenz [1]
Senior fellow in China Studies
Victims of Communism Memorial Foundation 


The “Karakax List”, named after the county of Karakax (Qaraqash) in Hotan Prefecture, represents the most recent leaked government document from Xinjiang. Over 137 pages, 667 data rows and the personal details of over 3,000 Uyghurs[2], this remarkable document presents the strongest evidence to date that Beijing is actively persecuting and punishing normal practices of traditional religious beliefs, in direct violation of its own constitution.

Specifically, the Karakax List outlines the reasons why 311 persons were interned and reveals the cognition behind the decision-making processes as to whether individuals can be released or not. Based on the principles of presumed guilt (rather than innocence) and assigning guilt through association, the state has developed a highly fine-tuned yet also very labor-intensive governance system whereby entire family circles are held hostage to their behavioral performance – jointly and as individuals. Ongoing mechanisms of appraisal and evaluation ensure high levels of acquiescence even when most detainees have been released from the camps.

The detailed new information provided by this document also allows us to develop a more fine-grained understanding of the ideological and administrative processes that preceded the internment campaign. In particular, this research paper carefully reviews the sequence and timing of events during Chen Quanguo’s first seven months in the region. It is argued that Chen must have been installed by the central government, possibly during a meeting at the Two Sessions in Beijing in March 2016 where Xi Jinping, Chen, and Chen’s predecessor in Xinjiang, Zhang Chunxian, were all in the same place. It is argued that Chen’s role in Xinjiang has not so much been that of an innovator as it has been that of a highly driven and disciplined administrator, with a focus on drastically upscaling existing mechanisms of investigation, categorization and internment.

More than any other government document pertaining to Beijing’s extralegal campaign of mass internment, the Karakax List lays bare the ideological and administrative micromechanics of a system of targeted cultural genocide that arguably rivals any similar attempt in the history of humanity. Driven by a deeply religio-phobic worldview, Beijing has embarked on a project that, ideologically, isn’t far from a medieval witch-hunt, yet is being executed with administrative perfectionism and iron discipline. Being distrustful of the true intentions of its minority citizens, the state has established a system of governance that fully substitutes trust with control. That, however, is also set to become its greatest long-term liability. Xinjiang’s mechanisms of governance are both labor-intensive and predicated upon highly unequal power structures that often run along and increase ethnic fault lines. The long-term ramifications of this arrangement for social stability and ethnic relations are impossible to predict.

Chinese document: download first twelve rows (redacted)

English document: download an English translation of the first 12 rows (redacted)

1.0 The Karakax List: Background and Overview

Source, Structure and Contents

Only days after the publication of the China Cables on November 24, 2019, Asiye Abdulaheb, who had previously leaked the China Cables, gave the author a document titled “Students sent to re-education [who are] family members of those who went abroad and have not returned” (出境未归人员亲属送培学员). This new file had been leaked by the same source from within Xinjiang as the other China Cables documents. The file was also given to a Uyghur in exile, who then gave it to a range of western media outlets. For the purpose of this research, this document is referred to as the Karakax List.

The leaked document is a 137-page PDF that was likely generated from an Excel spreadsheet or Word table. The reason for this assumption is that it is made up of spreadsheet or table-like columns and rows, except that some of the cells contain text that has been cut off. Alternatively, the original data was in the form of a database document, with data displayed in a table-style column and row format. In either case, the conversion to the PDF format was likely done for the purpose of leaking it outside the government system. This conversion was done unprofessionally and caused a (relatively minor) amount of data loss. Importantly, errors like these make it in fact more likely that the document is authentic as attempts to create a forgery would have aimed for a more perfect look.

The primary purpose of the data set contained in the Karakax List is to systematically present all relevant available evidence for each “main person” that is used to arrive at a verdict of whether each respective person should be released or not.

The document contains 667 rows, two of which encapsulate the same data. The resulting 665 individual data rows contain the unique identities of 311 main persons, with duplicate rows mainly differing between each other in terms of the final verdict that they render. These 311 main persons have the following in common: a) they are all from Karakax (Qaraqash) County in Hotan Prefecture, a region with Uyghur population shares over 90 percent; b) they are all related to persons who have gone abroad and not returned to China (出境未归人员); c) they all were at one point interned in what the state has called “Vocational Skills Education Training Centers” (职业技能教育培训中心), and what the author refers to as “Vocational Training Internment Camps” (VTICs); meaning, some of them have a verdict in the document that they were released but remain under close scrutiny, and may be re-interned; d) they all are (or were) interned in a VTIC in the Karakax County seat, as opposed to facilities in or closer to their home regions (townships); this is notable, because most (or perhaps all) townships have their own re-education facilities. At least in 2014, tougher “problem cases” were handed over to county seat camps, but it is not clear whether this practice remains in place.[3] Notably, b) is not the reason why most of them were interned in the first place. Those main persons, along with numerous other interned persons listed in the document, were interned for all sorts of reasons, with links to anything ‘abroad’ being a significant but not primary factor.

Figure 1. The first (redacted) page of the 137-page PDF. Source: Uyghur Human Rights Project.

Figure 2. Example of one row with two row numbers and cut-off text fields. The resulting data loss was most likely caused through the conversion of the original data into the PDF format.

The contents of the document is structured as follows. The fist 10 columns of each row contain information in relation to this person, including the location, date and reasons for their re-education internment. The 11th column contains the largest amount of text, describing each main person’s circle of relatives (亲属圈), community circle (社会圈) and an account of their “religious inheritance circle” (宗教传承圈), the latter referring to the ways in which religious knowledge was transmitted to the main person through (typically older) family members or other persons.

In total, the document lists 2,802 adult persons, besides hundreds of related minors. Of these 2,802, 1,432 are shown with their full names and ID numbers, and 1,370 only by name, age, and often address and/or occupation. Of those with ID numbers, all have IDs from Xinjiang, 1,405 have IDs from Hotan Prefecture, and 1,372 have IDs from Karakax County (locations of current registration are indicated by the first six digits of the ID). The total number of adults must be considered an approximation as it might contain several dozen duplicates that result from persons with only names and no ID numbers, where the document shows ages that only differ by one year. This may have resulted from the age of a person being recorded at different points in time. In addition, Uyghur names are phonetically transliterated into Chinese, which can result in several different legitimate character combinations. Duplicates for persons with ID numbers, including all 311 main persons, were checked and eliminated. The same was not done for those without ID numbers, given that the total number of distinct identities is only incidental to the analysis and verification of the document.

Table 1 shows that when disregarding the subsequent releases from re-education, 656 or 23.4 percent of the 2,802 persons are or were at one point in time interned or imprisoned.

Designation Count Share
Not interned 2,146
Sentenced (prison term)    137   4.9%
In detention center      42   1.5%
In re-education (at peak, disregarding subsequent release verdicts)    477 17.0%
TOTAL 2,802  
All forms of internment or prison (at peak point, subsequent disregarding release verdicts)    656 23.4%

Table 1. Note: The original number of persons in re-education was 484. The table shows a lower number since some of these were subsequently sentenced to prison terms, and therefore categorized as such in this data set.

The 12th column contains the “overall evaluation recommendation” (综合研判意见) or verdict issued by the main person’s local governing entity. In some instances, a different entity already arrived at a verdict, which is then often stated at the bottom of the 11th column, in which case the 12th column only contains an “agree” or “disagree”. These verdicts are essentially based on the totality of the information contained in each row, along with additional information that is often briefly cited in the verdict text. Most main person have more than one data row. A careful analysis of these rows shows a logical and chronological evolution of verdicts, which can change from “ongoing study” to “release”, vacillate between the two, or (much less commonly) take a turn for the worse.

Overall, nearly three-quarters (74.3 percent) of all main persons ultimately ended up with a “release” verdict. When including those who are now just required to finish the minimum one-year study period, the effective release share amounts to 87.5 percent. Release verdicts often state that a person first has to complete what is evidently a minimum one-year study period before they should then be released. It is important to keep in mind that “re-education internment” for all of these main persons exclusively refers to internment in VTICs. These camps will intern persons for designated periods of time before releasing them into either community control (社区管控) or, more commonly, forms of forced labor. In fact, a number of verdicts bluntly state that persons are to be released into “industrial park employment” (园区就业).

Throughout the document, quite a few of the first 10 data columns are left blank. In this regard, the repeated information in the document turns out to be a significant benefit, because when taken together, the author was almost always able to obtain a complete data set for each main person. Duplicate data sets also at times provide more detail, such as providing more detailed reasons why a person was interned in the first place.

1.2 Significance of the Document

During the past two years, the author has seen a lot of information that is similar to that contained in the Karakax List in various other documents. However, he has never seen them combined in the same document, and never observed verdicts and the reasoning behind them. As a result, the information provided in the Karakax List is truly unique in the public record, yielding powerful new insights about the internal anatomy of the internment campaign and the subsequent transition into forms of involuntary labor and community control.

In that sense, the Karakax List is the perfectly complementary third leak after the Xinjiang Papers (New York Times), which largely consisted of higher-level information such as speeches, and the China Cables (International Consortium of Investigative Journalists), whose most important document was a regulation (titled “telegram”) on how to run the Vocational Training Internment Camps.

This most recent leak provides the by far most detailed account of the inner decision-making dynamics of Xinjiang unprecedented campaign of mass internment. While confirming details from countless witness accounts, it reveals the witch-hunt like mindset that has been and continues to dominate social life in the region.

1.3 Dating the Document

The author has knowledge of the fact that the document must have leaked out of Xinjiang before late June 2019. Additional confirmation that the data is older than July 31st, 2019, comes from the fact that a file showing public service jobs, dated on that day, shows different employment designations than the Karakax List for a total of 13 persons that feature in both documents.[4] At the same time, the most recent date shown in the Karakax List itself is March 8, 28, 2019.

One female features in the Karakax List as a 37-year-old farmer who lives at home. In a secondary document that contains the identities of 59 persons in the leaked document, the same person is shown as being trained in the field of elderly care in the Karakax No.2 Farmers and Pastoralists Higher School of Technology (墨玉县二高农牧民技工学校). That training was scheduled to take place between March 2 and May 2, 2019. There are numerous other cases like hers, but with one notable exception. A 24-year old divorced woman was trained at an educational facility as a waitress between March 2 and May 2, 2019. The Karakax List shows her working as a waitress.

The most straightforward explanation for all this is that each data row shows information from a different date in time. Consequently, the age of each row’s information differs. Some of the rows show data that was current in early 2019, while others contain more recent information, including data entries that are more recent than May 2, 2019. The perhaps most likely date of the creation of the leaked PDF is close to the time that it was leaked out of Xinjiang, possibly in June 2019.

1.4 Authenticating the Document

Given that it is a data spreadsheet turned into a PDF, the Karakax List possesses no external markings such as stamps or signatures that could be used to verify its authenticity. Rather, such authentication must be based on its actual contents.

This report employs a three-step process to verify the authenticity of this document. First, the author verified the personal identities of 337 persons listed in the document through three other sources: a) a large cache of local Xinjiang government documents obtained by the author from different local government departments in minority regions (village to township levels), predominantly Hotan, Kashgar and Kizilsu Prefectures[5]; b) a leaked database with identity information from the Chinese surveillance company SenseNets; and c) spreadsheets and files obtained from Xinjiang government websites.

Second, the author verified locations mentioned in the document, most notably the locations of re-education camps. This will be discussed in section 2.0.

Third, the terminology used in the document was evaluated by comparing it to the terminology found in other government documents. This will be discussed in sections 3 and 4. However, a brief comparison of the Karakax List to the main document of the China Cables on how to operate VTICs yields a number of important data points, for example:

  • Both state that persons must be in VTICs for at least one year.
  • The VTIC management document (or “telegram”) of the China Cables state that the release of a VTIC detainee requires a multi-stage approval process, involving detailed reports and cross-checks with the Integrated Joint Operation Platform (IJOP) to see whether additional issues with them or their family members arose in the meantime. The Karakax List shows release verdict texts in succinct form, and also recommends people for release because “no other issues” have arisen.
  • Both reflect the fact that VTIC detainees are closely monitored after their release. The Karakax List indicates that persons receive approvals to remain released, meaning that they will be required to return to the VTIC if their performance does not meet the government’s expectations.
  • The China Cables telegram mentions three VTIC management areas: “regular” (普管区), strict (严管区) and “forceful” or “very strict” (强管区). The Karakax List has a separate column for these areas, with the exact same three area names.

Fourth, the author evaluated the internal consistency of the document (see section 4).

Overall, the document contains a relatively small numbers of mistakes, all of which can be attributed to human error. Originally, the data contained in the Karakax List was almost certainly entered by hand. For example, there are major variations in the ways that the re-education camps in which main persons are (or were) interned are named. Similarly, the spelling of local place names varies within the document, and often deviates from spelling conventions found on maps or other documents on the Internet. Given that Uyghur place names are phonetically transliterated into Chinese, which permits a range of possibilities, this indicates that the data was entered by a range of different local Uyghur government staff.

When considering that it contains personal information for over 2,000 persons and possesses a considerable degree of complexity, the Karakax List shows a high degree of internal consistency and data validity. Within the document, the data in the first 10 columns pertaining to main persons appears to be overall more consistent and reliable than the data in the 11th column.

2.0 Karakax County and its Re-Education Camps

2.1 General Overview

Karakax County is part of Hotan Prefecture. Both regions are predominantly Uyghur. In 2018, Hotan Prefecture had a population of 2.53 million (2.45 million of them Uyghur).[6] Karakax County’s population stood at 646,202 in 2017, with 631,541 of them being Uyghur.[7] If assuming the same population growth rate as Hotan, the county’s 2018 head count would be closer to about 633,000.

Karakax has 16 administrative units: four towns and 12 townships. All of them are found in the document as residence locations, frequently with different or inconsistent spelling. The reason for this is again the fact that the local Uyghur population, including the predominantly minority cadres, use different phonetic transliterations for what were originally Uyghur names. Searches on the Chinese internet and Baidu Maps showed that the spelling variations contained in the document are not uncommon.

2.2 Internment Facilities

The leaked document shows that all of the 311 main persons are or were in one of four re-education centers (第x教培中心 or 培训中心). Several of the other persons are also interned in these camps or in one of the county’s detention centers. According to a Karakax government website dated May 2018, the county has a total of five re-education centers.[8] This website uses the exact same abbreviations to describe their names and (abbreviated) locations, as well as showing the vice heads of these centers. All five centers have the same naming convention and there is no indication that their purpose differs. The author was able to identify three of the four centers mentioned in the document on satellite imagery. All of them were constructed between 2014/15 and early 2018, most likely in late 2016 or early 2017.

All three identified camps had their visible security features (tall razor wire fences around buildings and in courtyards, watchtowers) removed in the first half of 2019, with two of them having these features removed between April and June 2019. With a massive 2,074sqm armed police base, a 2,771sqm canteen, and eight large main buildings, the no. 1 center appears to be the largest of these.

Center number 1 2 3 4
Center name 第一培训中心 or 博斯坦库勒培训中心 (more rarely spelled 波斯坦酷勒培训中心), or 墨玉县1区培训中心 第二培训中心(火车站) or 火车站培训中心培训


第三培训中心(老职高)or县老职业学校教训中心 or老高职培训中心




Center location 博斯坦库勒工业园区

(Bosidan Industrial Park)

37°06’42.7″N 79°38’30.9″E

Exact location is unclear 墨玉县职业技术高级中学

Karakax Vocational and Technology Senior High School (see Baidu and Google Maps), 37°15’7.79″N 79°43’19.88″E

Likely on or near the grounds of the 墨玉县第二高级中学 (see Baidu Maps), 37°15’33.88″N 79°44’52.93″E
Persons in charge of the center (副主任) 阿卜力提普·阿不力克木 (Uyghur) and胡志伟 (Han) 艾斯凯尔·麦提努日 (Uyghur) and阿卜力米提·麦提如则 (Uyghur) 艾乃斯·阿卜杜艾尼 (Uyghur), 阿卜杜力提普·如则尼亚孜 (Uyghur), 萨拉麦提·努尔 (Uyghur) and蔡亚楼 (Han) 艾山江·艾合麦提 (Uyghur), 凯迪尔丁·艾依热提

(Uyghur), 图尔荪巴柯·哈依拜 (Uyghur), 排尔哈提·麦麦提托合提 (Uyghur)

Visible features on Google Earth Camp and detention center were constructed between 10/2014 and 1/2018. Tall surrounding wall, tall fences around buildings and in open spaces, watchtowers. Security features were removed between April 22, 2019 and June 21, 2019. n/a Constructed between 7/2015 and 1/2018. Tall surrounding wall, tall fences around buildings and in open spaces, watchtowers. Security features were removed between May 8, 2019 and June 21, 2019.


Constructed between 6/2015 and 4/2018. Tall surrounding wall, tall fences around buildings and in open spaces, watchtowers.

Security features were removed between January 28, 2019 and April 7, 2019.


Evidence from construction bids or other government documents Construction bid for the camp, featuring 8 buildings (corresponds to satellite images).[9]


Construction bid for seven factories for this center at a cost of 81.2 million RMB, 墨玉县博斯坦库勒扶贫车间(墨玉县教育培训中心+扶贫车间建设项目(1-7号厂房))[10]


Construction notice for a 2,074sqm armed police station and a 2,771sqm canteen in this center.[11]

墨玉县职业技能培训中心实训基地第二区项目建设及附属配套设施建设项目(火车站)[12] Construction bid for conversion of former Vocational and Technology Senior High School to an educational correction center: 墨玉县教育矫治中心改扩建建设项目(设计),内容:对原职业高中改造及新建教育矫治中心的设计,具体内容详见招标文件。[13]




Construction notice for re-education poverty alleviation workshops on the grounds of the No. 2 Senior Middle School:  墨玉县物流园、二高、特校教培中心+扶贫车间配套项目, 建设地点: 墨玉县物流园、二高、特校[16]


Table 2. Authentication of re-education camps in which main persons are interned. Source: Karakax List.

Large centers apparently consist of several “districts” (区). This can refer to different areas on the same general compound, as in the case of camp no. 4, which has an “A district” and a “B district” (A区 , B区). [17] Camp no. 1, the largest of the three identified, apparently, has 12 districts (博斯坦库勒培训中心12区, row 12). This can be confirmed from a county government financial document that outlines the issuance of new bonds to finance alterations on districts 1 to 12 of the camp no. 1, which is also referred to as the Karakax County Education and Training Center (墨玉县教育培训中心建设及值班室安检通道改造建设项目(第1–12区改造续建)). [18] Besides being located in specific districts, re-education detainees can be very precisely identified through their district (区), building (楼) and class (班). For example, the father of the main detainee in row 415 is shown to be in re-education in “district 12, building 6, class 18” (12区6楼18班).

Conversely, “training centers” appear to be assigned to administrative districts (社区), and at least in Karakax, many of them their own “training center”. Throughout the document, centers number 1 to 5 are also referred to as “no. x district training center” (第x区[教育]培训中心). In the case of camp no. 2, the equivalence of “no. 2 training center” and “no. 2 district training center” is established in the document, since both terms are used in the same phrase and in relation to the same re-education detainee (墨玉县第二区培训中心(火车站培训中心培训)elsewhere also called 二区火车站). Even a “no. 12 district training center” (12区教育培训中心) is mentioned. Indeed, Karakax County has had 12 districts for much of its recent history (1950 to 2014, coinciding with the number of its townships).[19] This naming convention is therefore most likely different from the districts within individual camps (such as the district no. 12 in camp no. 1 mentioned above). The document specifically mentions “district camps” numbers 1 to 6, 10 and 12. If accurate, this relationship between camps and administrative units would broadly confirm the author’s previous estimated that the total number of re-education camps in Xinjiang is roughly commensurate to the number of its administrative units at township and higher levels.

According to government documents, the construction and operation of these re-education facilities became a grave financial burden for the county. A county financial operation analysis report from April 2017, a time when many minorities were being interned, notes a severe funding gap for a number of stability maintenance operations. Among them, the “education and training centers’ were forecasted to cause a funding gap of 108.9 million RMB.[20] As a result, the county dramatically increased the issuance of new bonds: from 187.4 million in 2016 to 532 million in 2017, of which 90 million were designated for re-education facility construction.[21]

In 2018, Karakax increased its bond issuance volume to a stunning 2.48 billion RMB.[22] Of the 25 projects that were listed as part of this financing drive, eight were related to re-education camps. Specifically, the new debt was scheduled to finance the expansion of re-education camps numbers one to four, along with the construction and expansion of related factories and skills training bases. A county report dated January 26, 2019, provides a detailed discussion of the contents and training goals of one of these re-education related factory construction projects. [23]

Satellite imagery from Karakax shows at least one massive and highly securitized factory compound, consisting of several dozen blue-roofed, single-floor workshops.[24] Constructed between April and September 2018, this compound measures approximately 325,000sqm, and features surrounding walls, tall razor wire fences around each single individual building, and several watchtowers. Just north of this compound are large new residential areas that appear to be part of a 380 million RMB poverty-alleviation employment and resettlement plan.[25]

The leaked document also notes at least two detention centers (看守所) by name. The Bostan Industrial Park Detention Center (波斯坦库勒看守所) is located directly south of the no. 1 re-education camp, southeast of the industrial park (37° 6’36.50″N 79°38’22.21″E). It originally consisted of two buildings built between 10/2014 and 1/2018. Between 04/2018 and 06/2019, eight detention center buildings were added, along with several other types of buildings on an extended part of its compound.

The size of the county’s main and first detention center, the Moy County Detention Center (墨玉县看守所), is located at 37°16’36.48″N 79°44’46.52″E, and likewise increased dramatically. Construction of this facility started in late 2011 and was completed by April 2013. Between 6/2015 and 9/2018, the land area covered by its buildings roughly tripled.

This massive expansion of detention centers indicates a dual-pronged approach as Beijing moves the internment campaign into the next phase. While especially those in the VTICs are being “released” into forced labor, many others in extrajudicial internment are being sentenced to long prison terms.

3.0 New Evidence on the Genesis of Chen Quanguo’s Internment Drive

By providing us with detailed reasons for internment and subsequent verdicts for release or continued “study”, the Karakax List sheds new light on the internal and administrative dynamics behind the internment campaign. Specifically, there is a strong overlap between the data and decision-making processes behind internment and verdicts, and the investigative tasks and categories of the so-called village-based work teams (驻村工作队). This section seeks to shed new light on both the ideological and administrative processes that led up to the internment campaign, culminating in a more detailed explanation of the genesis and timing of the re-education drive overseen by Chen Quanguo during his first seven months in the region.

3.1 “Educational Guidance must be Tough”: The Inception of Re-Education to Combat “Expressions of Religious Extremism”

In May 2013, the XUAR released a document titled “Autonomous Region Party Committee document no. 11” (自治区党委11号文件).[26] This document sought to distinguish between “legitimate” and “extremist” expressions of religion. It is widely cited in the context of the early stages of the crackdown on so-called Uyghur separatism and the related strike hard and re-education campaigns.

Most notably, this document emphasizes that “the hand that strikes hard must be tough, [and] the hand that [provides] educational guidance must [also] be tough” (打击的一手要硬,教育疏导的一手也要硬). The timing of this statement closely corresponds to the earliest mention of “transformation through education” in the context of Xinjiang’s de-extremification (August 2013), exactly one year before the progressive institutionalization of re-education both in terms of administrative structures and dedicated transformation through education facilities.[27] Several sources confirm the close connection between document no. 11 and the onset of the initial phase of the re-education drive. A January 2014 government report from Tekes County mentions “transformation through education” in the same sentence as that document, denoting re-education as one of the means through which its directives are to be implemented.[28] Similarly, a December 2014 article combines a discussion of document no. 11 with a visit to one of the earliest dedicated re-education camps, the Kargilik County Transformation Through Education Center (教育转化培训中心).[29] Finally, one of the most comprehensive reports on pre-2017 re-education work, a January 2015 state media article on Yining County, cites document no. 11’s statement about educational guidance having to be tough in the very first sentence.[30]

In order to “control the expressions of religious extremism”, the document no. 11 proposed  “three simultaneous approaches”, consisting of establishing positive beliefs towards the Party (正信挤压), a confrontation (of “extremist beliefs”) with modern culture (文化对冲), and regulating (wrong beliefs) with the rule of law (法治约束).[31] It further stated that “the infected minority masses” should, through educational guidance, be led in the direction of secularization and a modern lifestyle. The latter is precisely what the re-education camps have been implementing, and certainly with a “tough hand”.

Overall, it is clear that document no. 11 played a major role in laying the ideological and administrative foundation for the evolving re-education drive. It bluntly advocates an intrusive and heavy-handed sociocultural re-engineering approach that is predicated upon “tough educational guidance” designed to promote secularization and weed out religious customs, practices and beliefs. Essentially, Chen Quanguo’s internment campaign took these directives and systematically implemented them throughout the entire region.

3.2 “Visit the People”: The Crucial Role of Village-Based Work Teams in Identifying Persons for Re-Education Internment

In February 2014, Xinjiang’s Party secretary Zhang Chunxian initiated a campaign to send 200,000 cadres into rural minority regions for the three years of 2014 to 2016.[32] This campaign was titled “Visit the People, Benefit the People, and Bring Together the Hearts of the People” (fang minqing, hui minsheng, ju minxin, 访民情、惠民生、聚民心), abbreviated as fanghuiju (访惠聚).[33] This village-based work team strategy sought to penetrate deeply into Uyghur communities and homes, making cadres live, work and eat with the people, with the primary aim of gathering intelligence. The campaign was to achieve complete coverage, “leaving no blank spaces”.[34] Together with other government documents, the Karakax List testifies to the fact that this campaign has and continues to form the backbone of the concurrent re-education drive, even more than previously thought.

Notably, Zhang Chunxian’s campaign was not novel. It was Chen Quanguo who in November 2011 during his rule in the Tibet Autonomous Region initiated a massive drive of sending 100,000 cadres in village-based work teams to Tibetan villages.[35] For Chen, these “village-based work teams” (驻村工作队) represented a key pillar of his government’s stability maintenance work, in addition to administrative and technological innovations such as mass surveillance systems or grid management. The work of these teams was extremely intrusive in nature, seeking to obtain personal information from Tibetan villagers that could be used to report on them.[36] Perhaps in order to confirm his original authorship of this initiative, the phrase “village-based work team” (驻村工作队) was added to fanghuiju as soon as Chen took over from Zhang.

When Chen Quanguo replaced Zhang Chunxian as Xinjiang’s Party secretary, the 3-year campaign was almost completed, and valuable lessons had been learned from it. Chen immediately placed a special emphasis on the village-based work teams, featuring them and their importance in the very first section of his speech on October 29, 2016.[37] When the village-based work team campaign was re-launched in 2017, it came with a mandate of special urgency. The campaign was to achieve a “standardization, scientific-ization, institutionalization, and normalization of fanghuiju village work”.[38] The teams represented not only the essence of grassroots work, but also a long-term mechanism for penetrating and permeating the grassroots.

3.3 Village-Based Work Teams and the Sequence of Events Leading up to the Internment Campaign

Importantly, the timing of the village-based work teams could have played a decisive role in the timing of the mass internment drive. Chen Quanguo was likely brought to Xinjiang in August 2016 so that he could spend the first 6 to 7 months on a massive securitization drive, involving mass recruitments of police forces, constructing thousands of new police stations and checkpoints, and installing large-scale surveillance systems. The establishment of this security infrastructure would have been a first move in order to prevent any possibility of public unrest resulting from the mass internments.

Meanwhile, Chen began to adapt and prepare the village-based work team campaign for the upcoming internment drive. In a November 16, 2016 speech, he made 10 new demands for this work. Rather than pointing to innovation, these demands reflected a planned evolution of the work team campaign, focusing on their work focus, effectiveness, complete coverage and iron-fisted discipline.[39] This culminated in the development of a regionwide new village-based work teams plan (2017年深入开展“访民情惠民生聚民心”驻村工作计划). Again, rather than featuring significant new contents, this plan primarily focused on the urgency and overarching importance of the teams’ work for maintaining social stability and penetrating grassroots work.[40] In a widely cited January 7 speech that was broadcast on television and listened to by all stability maintenance-related government work units (including the village-based work teams), Chen focused on the subject of Party discipline, emphasizing the need to “strictly manage the Party [using] iron-like discipline”.[41] The speech reiterated the need to conduct “penetrating fanghuiju village-based team work”, and echoed Xi Jinping’s words from 2014 that the region was to “deploy a net from heaven to earth [and] building copper wall and iron walls” (布下天罗地网、构筑铜墙铁壁).[42] Again, Chen’s main focus was on preparing his ranks for the tough-minded, military-style investigative and internment operations that were to follow. On January 20, the region then launched a new “fanghuiju village-based work team management information system”, a “quantitative analysis tool to undertake statistical analyses of the situation of villagers and villages” (数据分析工具进行村民村情统计分析). [43]

The launch date of the internment drive was likely not just dependent on the time needed for Chen to adjust to his new job, gain control over the state apparatus and securitize the region. Village-based work teams had established time windows for their annual re-launch (Table 3). The teams were formally sent-off with a regionwide ceremony and had set dates by which they had to be fully deployed. Aside from a minor deviation in 2016, these dates were in early March. For 2017, work teams were for the first time ordered not to leave during Chinese New Year. They were required to maintain a skeleton team presence through February 1st, with full team deployment to be completed by March 1st.[44] The internment campaign received a quasilegal backing with the issuing of the “de-extremification ordinance” on March 30. All other evidence, including the internment dates contained in the Karakax List, indicate that the campaign started in early April 2017 (see section 4.2.2).

Year Chinese New Year holiday period Dates of team send-off day / full deployment deadline
2014 Jan 31 – Feb 6 Feb 14 / March 5
2015 Feb 19 – 24 Jan 15 / March 1
2016 Feb 8 – 13 Feb 24 / Feb 25
2017 Jan 28 – Feb 2 Feb 25 / March 1

Table 3. Sources for send-off and deployment dates: 2014: (compare, 2015: and (compare and, 2016: and (compare, 2017: and

This sequence of events also indicates that Zhang Chunxian’s rule in Xinjiang was almost certainly cut short, so that Chen Quanguo could be installed in time to oversee what must be considered a premeditated internment campaign. Zhang’s signature campaign, the fanghuiju, was designed to finish in late 2016 with a general evaluation. Instead, this evaluation was overseen by Chen, more than likely in order to give him an opportunity to become acquainted with this work in Xinjiang, and to adapt it for the subsequent re-education drive. Chen Quanguo, Zhang Chunxian and Xi Jinping were all in the same place during the Two Sessions in Beijing in March that year, making it possible that detailed plans for Xinjiang’s internment campaign and related governance changes were made at that time.

3.4 “Digging Behind the Curtain”: The Investigative Mission of Village-based Work Teams

On February 2, 2017, approximately eight weeks before the onset of the massive internment campaign, the Hotan Prefecture Fanghuiju Village-Based Work Team Leadership Small Group Office (和田地区“访惠聚”驻村工作领导小组办公室) issued a set of work instructions from the prefecture’s Party secretary.[45] The document obtained by the author was stored in the file cache of the Töwet (Tuowaite) Township Fanghuiju village-based work group, which is in Karakax County. Each cadre, village police officer and village-based work team member was to be responsible for 40 households, ensuring full coverage by “spreading pressure” and roping in existing local arrangements such as neighborhood watch systems and linked household groups (a mechanism by which 10 households were mutually responsible for supervising and reporting on each other). Their work principle was to “go out in the morning to work, focus on visiting homes, and collect and evaluate work [results] in the evenings”.

More specifically, the teams were to adopt a strategy of employing nine different investigative techniques (九类工作法) and then to hold daily evaluation meetings where they would discuss the results of their “excavation work” (揭挖工作). The header for this section was titled: “Implement the overall goal, dig behind the curtain, shovel the soil” (落实总目标,挖幕后、铲土壤). It continued by mandating the establishment of a cohort (专班) dedicated to “dig the stock, reduce the amount, shovel the soil” (挖存量、减增量、铲土壤), or short: “dig, reduce, shovel” (挖减铲). The “dig, reduce, shovel” motto was initiated by Chen Quanguo during his crucial October 29, 2016, speech, where he advocated it in the context of establishing a “fine-grained strike-hard” approach.[46] This phrase refers to three aspects: a) to dig out and hence uncover the problem persons (“terrorists” and their helpers); b) to reduce their numbers; and c) to shovel away the soil where they can grow, flourish or hide.

This concept coined by Chen was arguably, again, an evolution rather than an innovation. In January 2015, Zhang Chunxian had coined the phrase “remove the cover, dig behind the curtain” (揭盖子、挖幕后), or short: “uncover and dig” (揭挖), as a vivid metaphor for uncovering the places of hiding and influence of extremist elements in the context of the “strike hard campaign” (严打暴恐专项行动).[47] A 2016 Hotan Prefecture spreadsheet designed to record the activities of village-based work teams (in this case for Karakax’s Töwet Township) featured a column where works teams have to enter how many times they “accompanied the ‘remove the cover, dig behind the curtain’ work” (配合开展”揭盖子、挖幕后”工作).[48]

Under Chen, both concepts came to be used in tandem. For example, a village-based work team report from July 2017 from Pishan County speaks of an “ ‘uncover and dig’ and ‘dig, reduce, shovel’ special activity” (“揭挖”“挖减铲”专项行动) in the context of investigating two particular problem groups: so-called “untrustworthy persons” (不放心人员), and persons from the village who could not be accounted for.[49]

Crucially, both of the terms “uncover and dig” and “dig, reduce, shovel” are directly associated with the investigative methods of the village-based work teams and how these methods dramatically contributed to the mass internment campaign, especially in the first half of 2017. Here are four pertinent examples.

  1. the village work team in Toghraqbostan (托格拉克博斯坦村) village in 拜什托格拉克 Township, Lop County, Hotan Prefecture, states that “based on notifications from higher [state] levels and the village-based work team’s ‘uncover and dig’ efforts, the number of strike-hard detainees rose from 6 at the beginning of the year [2017] to now 16, and the number of persons in education and training increased from 6 to 21”.[50] The village had 154 households.
  2. A very similar report from Qedirdan (喀迪尔墩村) village in the same township, dated March 2017, says that “since the start of the ‘dig, reduce, shovel’ [campaign], up until now, 23 persons were detained and sentenced, and 19 persons [placed into] transformation through education”.[51]
  3. 2017 Hotan CPC District Committee work report noted that as part of the “’uncover and dig’” initiative, 179 persons had been detained and 163 persons “sent to transformation through education training.”[52]
  4. A June 2017 report from the Uqturpan (Wushen) County Forestry Department Village Work Team noted that “through information evaluation meetings, 54 untrustworthy persons had been sent to the legal system school to study”.[53] These meetings are part of the investigative mechanism called for by these “dig” and “uncover” campaigns.

In tandem with village-based work teams, the “the dig, reduce, shovel” campaign was instrumental in the internment campaign. A document issued by the Arslan Bagh (阿尔斯兰巴格) Township government notes that this campaign is, along with other measures, being used to “cut off the extreme religious thought inheritance system and eradicate the … transmission channels of religious extreme thought” (切断宗教极端思想传承体系,铲除宗教极端思想源头和传播渠道).[54]

Notably, the “dig, reduce, shovel” concept features prominently in one of the leaked documents from the China Cables: bulletin no. 20, which features a report titled “Integrated Joint Operation Platform [IJOP] Daily Clues Report”.[55] In this document, these clues are referred to as “dig, reduce, shovel clues”.

The relationship between the village-based work teams and the IJOP is evidently circular:

  • Through the IJOP app (一体化APP) and other local reporting mechanisms, the teams, along with local cadres and government staff, feed the integrated big data platform with meticulous details about households and individuals based on their home stays and visits.
  • The IJOP combines all data from all sources, including police and surveillance data, and then flags anything that is deemed suspicious and warrants further investigation.[56]
  • Besides local cadres and police, it is often the work teams who are then charged with investigating whatever the IJOP flagged, included so-called “persons recommended to be sent [to re-education] by the integrated platform” (一体化平台推送的人员, or short: 一体化推送).

Specifically, work directives for these teams mandate that they must “undertake an on-the-ground investigation of micro clues [微线索] that were pushed [by the IJOP].”[57]

Together with the local authorities, village-based work teams are also responsible for comprehensive analysis of the evidence that they gather. All four examples mentioned above of the teams boosting the ranks of those sent to re-education mention information evaluation meetings (信息研判会) in more or less direct conjunction with sending persons to re-education.

After long days filled with home visits, the teams are then required to hold daily “evening evaluation meetings” (晚研判会), during which they examine the information that was gathered, evaluate whether it may contain elements of suspicion, and note the results in a report. An overview of work team duties from Hotan County notes that the evening meetings consider a range of information, including household data, their “actual behavior”, and other collected materials.[58] An example of such a report comes from Töwet Village in Karakax’s Töwet Township. Dated March 1, 2017, the report notes that the village has many persons who illegally crossed the border into other countries, including many who are now stranded overseas and have not returned (非法出境和合法出境非法滞留未归人员). [59] The report further speculates that there is a risk that those persons abroad may be secretly collaborating with local villagers to commit crimes. The region is considered to have a “dense religious atmosphere” (宗教氛围浓厚), and its location at a road junction, with related uncontrolled comings and goings, is noted to be of particular concern. Indeed, both a range of other, related documents as well as the Karakax List indicate that those two areas of concern played a crucial role in designating persons for re-education internment.

3.4 “Trustworthy” or “Untrustworthy”? The Investigative Categories of Village-Based Work Teams

There is a strong correlation between the terminology of reasons for internment found in the Karakax List and the investigative categories given to the work teams. A spreadsheet completed by the Yéngi Awat (英阿瓦提) township in Yarkand (Shache) County, Kashgar Prefecture, features a tab with a heading titled “remove the cover, eliminate the nest, dig behind the curtain activity survey results summary table” (“揭盖子、端窝子、挖幕后”活动调查情况汇总表).[60] This table (Figure 3) contains a detailed list of types of potentially suspicious or dangerous persons, including: religious figures (宗教人士); persons with no official religious duties who possess religious knowledge (有宗教知识没有职务人员); persons who went on a Hajj (pilgrimage to Mecca) – either in general, or in the form of a “scattered Hajj” (not as part of a government-approved tour group) ([零散]朝觐人员); persons who have been through re-education (已教育转化人员); men who in the past had long beards (以前留过大胡子人员); women who in the past wore veils (以前蒙过面的妇女); persons who work long-term outside their home region (长期外出务工人员); persons who sold their property and left [the region] (变卖家产外出人员); persons whose whereabouts are unknown (去向不明人员); and the arbitrary catch-all category of “persons in need of re-education” (需要教育转化人员). From the Karakax List, witness statement, and other documents, it is evident that these groups of persons were then prominently targeted for internment. Evidence for the link between the “dig, reduce, shovel” and the “remove and dig” campaigns, village-based work teams, and investigative categories such as “untrustworthy” or “focus” persons can also be found in numerous other sources and government websites.[61]

Figure 3. The “remove the cover, eliminate the nest, dig behind the curtain activity survey results summary table” for village-based work teams. Source: Source reference code: 34912.

This investigative spreadsheet also indicates that the work teams were expected to proceed in a very systematic fashion. In one tab, they had to compile a detailed record of the religious situation in the village. This included the number of places of worship, number of religious books, and many of the types of persons mentioned above (Hajj, religious knowledge, etc.), including women who wash their bodies according to religious rites. In a second tab, work teams were to record demographic details, with a special focus on age groups that were a special target of the re-education campaign (Figure 4). Specifically, persons were to be categorized according to decade in which they were born, such as “80s and after” (80后), “90s and after”(90后), and “2000s and after” (00后), separated by gender. All three of these demographic designations are found in the Karakax List as reasons for being sent to re-education. Finally, a third tab (Figure 4) was to be used to categorize persons as either “focus persons” (重点人), “average persons” (一般人员)  or “trustworthy persons” (放心人). This scheme existed prior to Chen Quanguo and is for example mentioned in a Karakax County village-based work team document from March 30, 2016, where teams are exhorted to categorize local households according to a four-fold scheme: two different types of “focus households” (重点工作户 or 重点管控户), “basic [average] households” (基本户)  and “trustworthy households” (放心户). [62]

Figure 4a. Persons categorized by perceived threat level and by the decade in which they were born. Source: Source reference code: 34912.

Figure 4b. Persons categorized by perceived threat level and by the decade in which they were born. Source: Source reference code: 34912.

These categorization schemes are of particular importance for understanding the dynamics of the internment campaign. Prior to the onset of Chen’s massive internment campaign, they focused on identifying “strike-hard detainees” along with persons to be subjected to in-situ ideological (re-) education measures. As the internment campaign unfolded, especially the softer “untrustworthy” category became a prominent tool for identifying re-education internment targets.[63] While this designation had previously often (though not always) been more of a general category that could encompass other problem groups, its use under Chen became more specific for identifying soft re-education (versus “strike-hard”) targets.[64] This is also how the term is ubiquitously used in the Karakax List.

Under Chen, the division of villages, households or persons during the course of the mass internment campaign involves up to four different categories. The most comprehensive scheme appears to involve four of them. The two problem categories are “focus persons” (重点人员) and “untrustworthy persons” (不放心人员), frequently also involving a third problem category of “special groups” (特殊群体).[65] These stand in contrast to “average persons” (一般人员) and “trustworthy persons” (放心人员). [66] This classification scheme also has consequences for how work team visits must be conducted. Focus persons and untrustworthy persons require that work teams are accompanied by police forces, and they are visited by a larger team of 7-8 persons. In contrast, the other two types of households do not require any police accompaniment and are visited by teams composed of 4-5 persons.[67] A Yarkand (Shache) County handbook for establishing regular village work mandates that wireless video surveillance systems must not only be placed at key locations and road intersections, but also near the homes of “untrustworthy persons”.[68]

Specifically, “focus persons” appear to be more likely to end up in detention (and ultimately, imprisonment), whereas “untrustworthy persons” appear to be mainly sent to re-education facilities (if they are not kept under local community control). This is confirmed by two of the reports cited in section 3.3. Both the Uqturpan (Wushen) County Forestry Department Village Work Team report and the 2017 Hotan CPC District Committee work report state that “untrustworthy” persons were sent to re-education facilities, whereas “members of terrorist groups and focus persons” were sent to detention facilities.[69] This distinction is also supported by other documents. For example, the 2018 Qargilik (Ruoqiang) County Justice System budget report states that “untrustworthy persons” are placed into social correction and drug rehabilitation centers (社区矫治康复中心), alongside drug addicts and mentally ill persons, where they receive a “new life” through “psychological rehabilitation, behavioral correction, and a reconfiguration of their value systems” (心理康复、行为矫治、价值观重建).[70] This represents a form of extrajudicial internment, akin to re-education, and not formal imprisonment. Finally, a spreadsheet from Kawak (喀瓦克) Township in the northern part of Karakax County that lists 1,249 persons in detention and re-education shows that 93.2 percent of persons who are classified as “untrustworthy” are either in detention or re-education.[71]

3.5 “Untrustworthy Persons” in the Karakax List

In the Karakax List, of 484 persons who are shown to be in re-education along with a specific reason, 116 (24.0 percent) were explicitly categorized as “untrustworthy person”, although it seems likely that this designation was not applied consistently. Of these 116, a total of 88 had this as the effectively only reason for internment, at times with minor elaborations, but not categorized for other main reasons. In contrast, only three of the 484 were labeled as “focus persons”.

Most commonly, “untrustworthy persons” were simply described as a general demographic category, following the “born after decade xy” scheme discussed above. Of the 116 persons with this designation, 98 followed such a demographic scheme. At other times, “untrustworthy persons” were simply deemed suspicious by association, for example by being part of the same family network as another person in detention. Others were classified as “untrustworthy” because they work outside of Karakax and only infrequently return home. Others had exceeded the official birth quotas or not allowed the three-year minimum time period between births. Yet another was said to “not go along with community work”, such as failing to attend the flag raising ceremony, while yet another was simply characterized as being “disassociated from society” (游离于社会面).

“Untrustworthiness” represents a general category of suspicion that is hard to grasp. It represents persons whom the state feels cannot be as easily understood or controlled as it wants them to be. A pertinent example is a “born in the 80s or after untrustworthy person” who has extensive business dealings outside the region. The “reason for re-education” states that this person has “relatively complicated interactions and activity trajectories”, and their whereabouts are “not easily grasped”. Several other cases in the leaked document give similar reasons for internment without designating those persons as “untrustworthy”, indicating that the category was not necessarily consistently applied, but often rather used as a catch-all reason for internment where little or no other real evidence existed. A particular pertinent example for this can be found in the release verdict text of an “untrustworthy person born in the 90s”. [72] Besides that designation, no other reasons for internment are given. The verdict simply states that “this person does not have [any] extreme religious thinking”, a blatant admission that is consistent with the generic and essentially totally arbitrary nature of the “untrustworthy” designation.

Consequently, “untrustworthiness” is the perhaps most pertinent mark of Chen Quanguo’s mass internment campaign, which, in contrast to his predecessor, aimed to deal with those hard-to-control population groups that were not easily amenable to criminal conviction, nor secular or steady enough to be considered “trustworthy”. The use of this catch-all category allowed the state to designate entire demographic groups as potentially suspicious, and hence liable for the “preventative treatment” of “transformation through education”.

3.6 Conclusions: Chen the Upscaler

The evidence presented in this section points to several new conclusions regarding the inception and unfolding of Xinjiang’s campaign of mass internment.

First, it appears to be increasingly clear that Chen Quanguo’s role in Xinjiang’s mass internment drive was less one of innovation than of execution. His main role was to adjust, optimize and especially upscale existing frameworks and mechanisms. In August 2016, Xinjiang already had an increasingly well-developed re-education framework: re-education procedures, mechanisms for identifying persons to be sent to re-education, and a system of dedicated (three-tiered) related institutions all existed.

Second, the internment campaign and the related replacement of Zhang with Chen was almost certainly premeditated. It must have been planned by a group of people, likely involving Zhang, Chen, and high-level central government figures. At least some of the face-to-face meetings may have taken place during the Two Sessions in March 2016 in Beijing. It is unclear who first thought of the mass internment campaign. It might have been Chen, based on resulting long-term outcome requirements set by Beijing. Two things, however, are clear. The overall implementation of this campaign involved the central government, at some point almost certainly Xi Jinping himself. And, when Chen assumed his new post in Xinjiang in late August 2016, he executed a premeditated plan. He immediately unfurled a rapid series of drastic and large-scale measures, knowing that he only had about seven months.

Chen first set up the security infrastructure (surveillance and police) in case the internment drive would create massive upheavals, and also to increase the flow of surveillance data that could be used for internment decisions. Secondly, he prepared and then re-launched the village-based work team campaign, which formed the groundwork for local internment decisions. Thirdly, he promoted a taught disciplinary spirit among the cadres, preparing them for what was to come. Even so, some of them later resisted the mass internment, and Yarkand County’s Party secretary Wang Yongzhi even released thousands of detainees.[73] Fourthly, he set the stage for a dramatic expansion of preschool education and boarding facilities, knowing that hundreds of thousands of parents would be interned. Finally, Chen oversaw the construction of a vast array of new internment facilities, evidently a highly secret campaign for which we have almost no data. A handful of public construction bids, analyzed by the author in his 2018 research paper, indicate that most of these notices were issued from March 2017 onwards, with a peak in June that year.[74]

None of these campaigns represented in and of themselves drastic innovations. What united them all was that they were executed with the same military-style drivenness and relentless precision.

Given all this, Chen was clearly the ideal man for the job. He not only possessed the experience with cutting edge police and surveillance mechanisms and technologies. He also had that uncompromising, intense drivenness and stern, military-like discipline required to upscale existing systems and mechanisms within extremely short time periods, creating vast new realities before the world even knew what was going on. Perhaps most importantly, Chen knew how to rule and manage less-than-disciplined minority cadres with an iron grip. He was familiar with innovation, but in Xinjiang he did not get caught up in experimenting with the new. Rather, he oversaw what we might liken to a wartime army-style operation, the mobilization of all levers of government, enabled by a shear limitless authority over human beings and unprecedentedly large financial budgets.

4.0 The Karakax List: Analysis and Authentication 

4.1 Verifying the Identity of Individuals

For the purpose of verifying the personal identities of those listed in the document, the author benefited from being in possession of a local government cache of over 25,000 files.  Some of these data sets come directly from locations in Karakax County. Many of these files contain spreadsheets with long lists of local citizens, and they feature detailed personal information.

The author recursively parsed these documents with a command line script, resulting in a list of 2.9 million unique ID numbers of persons from Xinjiang (ID numbers that start with the two digits ‘65’). Of these, 1.16 million are from Hotan Prefecture (ID numbers with the first four digits ‘6532’), and 255,189 from Karakax County (ID numbers with the first six digits ‘653222’). Of these, 321 could be matched with IDs from the leaked documents.

In addition, the author was able to obtain a list of ID numbers from persons in Xinjiang from an anonymous source in The Netherlands. This list derived from an unsecured online database of SenseNets, a Shenzhen facial recognition company that processes surveillance data.[75] The entire SenseNets database reportedly contained the identities of 2.5 million persons from Xinjiang. The list obtained by the author contains the ID numbers and names of 725,804 persons. Of these, 725,499 are from Xinjiang, 498,725 are from Hotan Prefecture, and 28,117 are specifically from Karakax County. A total of 33 IDs could be matched with IDs from the leaked document, of which nine had already been matched through the cache of local documents.

Finally, the author identified 577,413 ID numbers from persons in Xinjiang from documents, spreadsheets and web pages from a range of Xinjiang government websites. Of these, 58,158 are from Hotan Prefecture, and 31,843 are specifically from Karakax County. Five of them could be matched with IDs from the leaked document, with one of them having already been matched through the cache of local documents.

The personal identities of these 349 matching IDs were further verified by comparing names, and, where applicable, addresses and other personal details such as current education or employment with the other source documents. Additionally, the general validity (internal consistency) of each ID number was verified by running it through a Chinese ID verification website.[76] This resulted in a final count of 337 successfully matched and 12 unsuccessfully matched identities.

The matching process faced several challenges. Firstly, they arose from the fact that places of residence and occupation are subject to changes as the government requires persons to: relocate, enter training with the aim of changing occupation, intern, or be released from internment into old or new places of residence and employment. Since the secondary documents used to corroborate the Karakax Spreadsheet were published at different times (mostly 2017 or 2018), this issue was all the more pertinent.

Secondly, name inconsistencies are frequent. Since Uyghur personal and place names are phonetically transliterated in Chinese, there can be considerable, legitimate spelling variations. In addition, Uyghur names are often localized variations of Arabic names such as Mohammed, which means that names can have longer or shorter variants (e.g. Memet / Memetqurban), intimate or more formal variants (e.g. Turdush / Turdi), or other slight variations (e.g. Kerim / Abdukerim). In addition, incorrect names associated with ID numbers appear to be a recurring issue in local government documents. Such files also on occasion list the wrong gender of a person, contradicting the gender indicated by both name and ID number. This is all very likely a result of human error due to the immensely high reporting requirements and other stress factors that local government staff face.

The minimum matching requirements for the purpose of this research were as follows:

  • Residence and/or occupation can differ, but ID number and name must be identical
  • If either first or last names differ, then ID number, the other name part, and either residence or occupation must match.

These matching criteria resulted in the exclusion of 12 matches from the list, resulting in a final number of 337 successfully matched identities.

The author identified a number of other data issues that are likewise attributable to human error. Of the 1,815 ID numbers contained in the document, 29 have an incorrect length: two have 16 digits, 16 have 17 digits, and 11 have 19 digits. This represents an error rate of 1.6 percent. Notably, all of these numbers are found in the middle column (family and social circles), with none of them pertaining to the 311 main persons. Generally, it appears that this column contains more human errors than the other (and especially the main) data columns. The omission or addition of a final number in an 18-digit number sequence is clearly attributable to human error, resulting from manual entry. In all of these cases, the date of birth digits were valid. For six of the 19-digit ID numbers, the error stemmed from adding an extra digit. Omitting the 19th digit resulted in a valid ID number (per checksum verification).[77] In the other five cases, the 18th (checksum) digit was also incorrect.

Then, on two occasions two different identities were associated with the same ID number.  One ID occurred five times in the document and was associated four times with one identity (name / residence / occupation), each as a “neighbor” in the context of the same main person, and once with another identity and as part of the social circle of a different main person. Consequently, the source of this error is apparently a mix-up between different datasets associated with different re-education detainees. Another ID was associated six times with one identity, and three times with another. The root issue was identical with the previous case: the identity that occurred six times was the main person himself, while the one that occurred three times was the father of a (different) main person. A possible explanation for these errors is, again, that the main row data appears to be more accurate and reliable than the data for family and social circles.

Other instances where information for the same person differs between data rows can be explained by the chronological sequence of the data, meaning for example that a person who is shown to be interned in one data row is said to have been released in another.

Another apparent (but likely not real) discrepancy arises from the fact that a person may be shown to have been interned on a certain date when listed as the family member of a main person, but show a different internment date in his/her own main data row. In addition, main persons can have different internment camp names or numbers in different rows. Both of these issues can be explained by the fact that interned persons may be transferred between camps and hence have different (new) internment dates or camp designations.

These findings have significant implications for the authenticity of the document. If a person were to create a fake document solely based on knowledge of the components of Chinese ID numbers, meaning they would compute a random set of valid ID numbers for Karakax County, their chance of creating 337 ID numbers that actually matched those of real persons can be computed as follows. There would be 11.12 million valid ID number combinations: 33,480 (valid date of births for ages 0 to 90) x 2 (gender) x 166 (sequential code, represented by digits 15 to 17).[78] Given that the Uyghur population of Karakax in 2017 stood at 631,541, that means that it takes 17.6 attempts to generate a number that matches that of a real person. The result of the probability (p) of someone doing this 337 times is infinitesimal:

p=(631,541 / 11,115,360) ^ 337 = 4.086e-420

Therefore, the creation of the Karakax List with fake or invented data would only have been possible if that person had access to a sufficient number of real ID numbers. In addition, this person would have needed considerable insider knowledge of the terminology and the policy procedures of the internment campaign. It is far more probable that the data is real.

4.2 Analysis of the Interned Persons Sample

4.2.1 Demographic Composition

The author compiled a spreadsheet of all identities who were either main persons or shown as “interned for re-education” (送培 etc.) elsewhere in the document. This resulted in a sheet containing 484 unique identities: 311 main persons and 173 other persons in re-education.

Figure 5 shows the demographic composition of all of these 484 persons. This distribution is very similar to the author’s previous research findings.[79] Compared to the 2010 census, the age cohorts between age 25 and 49 years are overrepresented, indicating that these age groups have been particularly targeted. Interestingly, this age range exceeds that of the generic demographic categories of “untrustworthy persons”, which starts at “born after the 80s”, and therefore spans between the adult ages of 18 and 39 (when taking 2019 as the benchmark year). Compared to that range, the main age range of those interned as per the Karakax List is almost exactly 10 years higher (25 to 49 years versus 18 to 39 years). The author’s previous work noted that many interned persons are household heads, or generally those with influence in family and society. [80] The analysis below shows that reasons for internment vary with age cohort.

Figure 5. Source: Karakax List.

As in previous internment samples analyzed by the author, this sample is again predominantly male. Males made up 90.7 percent of interned main persons, and 90.0 percent of all interned persons.

4.2.2 The Chronology of Internment

Of the 484 persons shown as interned for re-education, 320 entries showed the dates of their internment. The results (Figure 6) strongly correlate with previous research findings that suggest that the internment campaign started in late March and early April 2017. Nearly half (44.1 percent) were interned in the three months of April, May, and June 2017. About two-thirds (67.2 percent) were interned in 2017, and about one-third in 2018, largely due to a peak internment intake in May 2018. Bulletin no. 14 of the China Cables noted that in the week from June 19 to 25, 2017, 15,683 “suspicious persons” in four minority population dominated regions in southern Xinjiang were interned as a result of IJOP notifications. The Karakax List would indicate that this period was indeed a peak season of internment (and therefore likely not representative of subsequent time periods).[81]

Figure 6. Source: Karakax List.

The second peak of internments between March and June 2018, during which 26.6 percent of all internments in the document took place, is potentially related to another round of village-based work teams in the spring of that year. It also coincides with a drastic increase in the floor spaces of several existing re-education facilities.[82]

This means that nearly all would have completed at least one year of internment by late June 2019, the latest possible leak date of the document, and when three camps identified on satellite imagery had removed most or all of their security features.

4.2.3 Reasons for Internment

After evaluating all relevant data rows, only 12 of the 484 re-education persons had no recorded reason for their internment. The reasons for the other 472 identities were coded into eight categories. The coding was not mutually exclusive, but one set of reasons could in theory result in an assignment in all eight coding categories. The coding category “link to anything ‘abroad’” includes: applied for a passport but never left, left the country and returned, was in contact with persons abroad, or has family members or relatives abroad.

The results are presented in figure 7. No single coding category was dominant, even though violations of birth control policies constituted the most commonly cited reason for re-education, often along with other (typically religion-related) reasons.

Figure 7. Source: Karakax List.

The fact that not one single category dominates the field can be explained by the preventative nature of re-education. Re-education focuses on persons whom the government deems to be of risk. They may have committed so-called “minor crimes” in the sense of behaviors that, according to the state, will likely develop into more serious issues, or indicate that a person may be susceptible to temptation to delve deeper into “religious extremism”. Overall, this perceived risk largely derives from what the state considers to be unknown factors: persons who are somewhat “religious” and therefore might become more religious, persons who live in families where members have been detained or convicted of crimes, or persons who associate with potentially “dangerous” or “suspicious’” people. Again, those with connections abroad are also classified as “risk” persons, firstly because of the potential presence of religious extremism, but more generally because those who live abroad are simply much harder to control.

Ultimately, the Karakax List demonstrates the extent to which Xinjiang’s re-education internment drive is a campaign to eliminate risk by achieving complete state control over individuals, families, communities, and regions. The state’s main mechanism for this is guilt by association.

Beijing has justified this campaign by arguing that those in the camps have committed “minor” or what we might term pre-crimes: persons who are on the verge of slipping into much more dangerous territory. The data from the leaked document soundly disproves these claims. The listed reasons for internment range from vagueness such as “untrustworthy person born in a certain decade”, to someone with a “minor religious infection” (受宗教极端思想感染轻微人员), or those who “by clicking on a website unintentionally landed on a foreign website” (点击网站链接无意登陆境外网站). Others were interned because their “thinking is hard to grasp” (思想难掌握), they had a “complicated network of relationships” (人际交往复杂), “petition other persons without reasons” (无正当理由缠访闹访人员), is “disassociated from society” (游离于社会面), “applied for a passport but did not actually leave the country” (办理护照未出境), merely has “relatives abroad” (境外有亲属), only “suspected of having watched downloaded terrorist or other religious extremist videos” (具有观看下载传播暴恐音视频或其他类宗教极端音视频的嫌疑), has “talked to persons overseas” (境外通联), used to wear a veil many years ago (such as e.g. from 2012 to 2014), used to grow a long beard years ago (in one instance between 2010 and 2014), used a phone with a number that is not registered to the person’s name, or merely “communicated with a detained person”. Often, these little “sins” are lumped together to justify internment. For example, a person may have traveled overseas and used to grow a beard.

Notably, one of the state’s strategies is to dig up the “sins” of the past. Those who used to veil or grow beards years before their internment did engage in a “religious” practice and are hence deemed more easily infectible by the “virus” of “religion”. But this method of the state likely has another purpose: it communicates to the victims of what anthropologists and Xinjiang academics Darren Byler and Joanne Smith Finley term “state terror” that the government’s reach is long.[83] Time cannot cover up what the state disapproves of because its reach is long and deep, penetrating into every household and digging intoturning upside down every pocket of privacy.

Overall, the Karakax List reveals a mindset whose perhaps most notable feature is the endless vicious cycles of suspicion by association, which then passes on to others who associate with that person. This is precisely the reason why, after her last visit to the region, Smith Finley wrote a piece titled “Now we don’t talk anymore”.[84]

Figures 8 to 10 show reasons for internment broken down by age cohort, with the percentages showing the relative share of the occurrence of a feature across all age cohorts. For example, the age cohort 25 to 29 years constituted 31 percent of all occurrences of persons interned for being “untrustworthy” across all age cohorts. The findings support the fact that “untrustworthiness” is indeed an age-related designation, largely limited to persons below the age of 40 (born in the 80s and after).

Figure 8. Source: Karakax List.

Figure 9. Source: Karakax List.

Figure 10. Source: Karakax List.

The same breakdown for religion-related reasons shows almost the inverse picture. Those aged 45 and above are more likely to have been interned for that reason than not, although religion-related internment was also a significant factor for anyone above age 25.

Finally, internment due to any form of connection abroad was most predominant for those over the age of 55, likely because they have the means to travel, but also because that age group is more inclined to travel for religious reasons (join a Hajj). Even so, this category featured fairly prominently among all age groups.

4.2.4 Verifying Reasons for Internment in Light of Policy and Other Sources

At the end of 2014, Xinjiang issued a document titled “Foundational knowledge for distinguishing religious extremism – 75 types of specific expressions” (识别宗教极端活动(75种具体表现)基础知识). The publication of this list was linked back to the “document no. 11” on distinguishing the normal from extreme forms of religious practice, and encouraged citizens to report any sightings of such behaviors to the police.[85] This list was again emphasized in numerous government websites in the spring of 2017, often in the context of the village-work team campaign.[86] The contents of this list form an important backdrop for understanding the reasons for internment stated in the Karakax List.

However, the reasons stated there are often very vague. Not uncommonly, persons were “sent to re-education because they were infected by extreme religious thoughts” (因受宗教极端思想感染送培), or simply “were influenced by extremist thinking” (受极端思想影响). Moreover, persons are listed to have been interned for reasons that are not in fact part of these “75 signs of religious extremism”, for example, for having had religious marriage rites (念尼卡, Arabic “Nikah”), or for “passionately donating money to a mosque” (给清真寺积极捐款). One person, whose verdict is not release but “continued study”, has a verdict statement noting that this person in 2014 donated 5,000 RMB to a local public mosque in Karakax township, as if such an act constituted a crime.[87] This confirms a report by Radio Free Asia, which stated that the Imam of a mosque in Xinyuan County was detained for refusing to provide the authorities with a list of persons who had made donations for the construction of a mosque.[88]

The other main set of reasons for internment, a connection to anything related to things outside the country, is often even more mystifying. Persons are not only interned because they visited one of the mostly Muslim “26 focus countries”, a practice reported by witnesses and previous media reports.[89] A substantial number of those on the list were interned simply because they applied for a passport, without ever leaving the country. Those who committed this “transgression” even carry a special designation in the document: they are referred to as “persons who applied for a passport without leaving the country” (办护照未出境人员). In one case where this is cited as the only reason for internment, it even comes with the blatant admission that there was “nothing unusual” about that person (办护照未出境无异常人员).[90] Some of those for whom this was listed as a reason for internment had applied for their passports several years before the internment campaign. Sections that discuss release verdicts frequently number those within a main person’s family circle who are abroad or who have applied for a passport, typically if these numbers are relatively high. In those cases, the verdict within that particular data row is often: “continued study”.

Within three months after taking up office in Xinjiang, Chen Quanguo had the passports of minority citizens confiscated, evidently to prevent them from traveling abroad.[91] A blunt admission that persons who simply applied for passports are in fact targeted by state comes from China’s Ministry of Commerce. One of the village-based work teams that was made up by members of its Xinjiang staff reported that 85 members of “special groups” received special training in “resisting extreme thinking”, including “focus persons” (重点人员), women who wash bodies of the dead according to religious rites, religious specialists, persons with religious knowledge who do not hold a related official position, and “persons who applied for passports”.[92] In a local list contained in the file cache of the Töwet Township village-based work teams, which is used to record all types of suspicious or “problem” persons, the names of people who applied for a passports are to be recorded in a section titled “Focus / special persons”.[93] That section is also used to list persons who are members of terrorist groups (including ISIS), “dangerous” persons, persons associated with the 2009 Urumqi Riots, persons who attended secret scripture study sessions, and persons who have already gone through re-education. The section about passport and travel is divided into three categories: 1) persons who applied for a passport, 2) persons who hold a passport but have not left the country, and 3) persons who left by illegally crossing the border.

Figure 11: Source: Source reference code: 19422.

A government report from Turpan City about village-based work teams clarifies that the mandate to closely observe suspicious persons pertains not only to “persons who went abroad and have not returned.” [94] Among those who are suspect are also those who have returned, as they are now classified as “flow-back persons” (回流人员). On several occasions, the Karakax List notes that persons were sent to re-education because they are “prevent flow-back persons” (防回流人员). Those returning from other countries, especially Muslim nations, should be prevented from “slipping back” unnoticed, because they might pose a danger to society, influencing those around them with the “extremist” ideologies or subversive strategies that they supposedly acquired. In order to “prevent” them from returning and freely going about their lives, the “prevent flow-back persons” approach places them under close surveillance. The county of Kargilik (Ruoqiang) even issued a purchase bid for 200 surveillance cameras, to be installed near the homes of “focus persons” and “prevent flow-back persons”.[95]

Attempted international travel is not the only issue that the state has with minority citizens. Anyone who often travels outside their home region is suspect. For example, one person was interned because he is a long-distance taxi driver, often leaves the region for substantial amounts of time, and communicates with many people. As a result, he is considered an “untrustworthy” person. Other, similar examples include persons who work in other parts of the country and are typically away from their home region. One of Chen Quanguo’s early policies in late 2016 was to require Uyghurs to return to their original places of registration and to greatly impede their ability to travel. According to reports, locals had to get official permission just to visit relatives in nearby villages.[96] Similarly, Uyghurs residing in other parts of China were forced to return to their native regions in Xinjiang. Ultimately, all of these measures are implemented by the state in order to gain complete control over people’s lives and movements. Preventing the free flow of persons helps the state map networks of relations and interactions, which, due to the principle of guilt by association, constitutes a key component of decisions for internment.

4.3 Family Circle, Social Circle and Religious Inheritance Circle

The data contained in the 11th column of the Karakax List show how the state seeks to evaluate individual behavior and thought patterns in light of their family and social contexts. The principle of “guilt by association” is driven by detailed evaluations of the numbers of persons in someone’s wider family context who are in some form of internment, and of how religious knowledge was spread in the family.

The 11th data column first lists the number of persons in a main person’s circle of relatives (亲属圈) and how many of them are in re-education or detention centers. This circle includes parents (including parents-in-law), spouses, children, and siblings. Besides showing name, ID number, residence and occupation, the spreadsheet lists their “current actual behavior” and notes if (and when) they were interned. If family members of a main person are interned or in prison, that fact counts against them in the verdict text. In one case where the verdict is “continued re-education”, the text notes that “many family members are in strike-hard detention” (家庭中多人被严打收押). That person has seven listed family members, six of whom are in detention or re-education.[97] Conversely, verdicts may contain the positive note that “among three generations of family members there are no persons in detention or re-education” (三代亲属中无严打收押教育培训人员).

The second data entry in the 11th column lists the “community circle” (社会圈) of a main person. Alternative names for this type of data are “neighborhood circles” (邻居圈) or “social circles” (社交圈). These typically list a few neighbors or friends, but often the entries only summarize how many within these circles are themselves in detention. Again, a main purpose of this type of analysis is to establish the extent to which main persons associate with others who are in a detention or re-education facility.

The final type of data entry found in the 11th column is titled “religious inheritance circle” (宗教传承圈). This data details the ways in which religious knowledge was transmitted to the main person through (typically older) family members or other persons. It gives details of religious habits, such as how often a person prays each day, whether they pray at mealtimes, go to the mosque, fast during Ramadan, or exhibit other forms of religious practice. Importantly, all of the religious practices listed in these sections are perfectly customary, peaceful, and appropriate expressions of Islamic religiosity outside of China. The way that they are presented in the Karakax List shows how the state treats them as illegal, a basis for discrimination, or at the very least problematic activities that must be closely examined and are used to arrive at negative verdicts against persons. If anything, this type of data presents the strongest evidence seen by the author that Beijing is actively persecuting and punishing normal practices of traditional religious beliefs, in direct violation of its own constitution.

The Chinese terms cited above are not uncommon in government documents, with the exception of “religious inheritance circle”. This term is rarely found, possibly due to its sensitive nature and specialized use. A notice issued by the Karakax County Party Committee in April 2017, which mandates a careful investigation of local cadres, states that this investigation is to evaluate “circles of relatives, religious inheritance circles, and social networks” (家属亲友圈、宗教传承圈、社交网络圈).[98] The same wording is also found in a notice issued by the governments of Aksu Konaxeher (Wensu) County (Aksu Prefecture) and Karamay City.[99]

4.4 Verdicts

The final verdicts for main persons were evaluated by sorting the rows for each identity ascending by row number. The final row number contained the final verdict as shown in the document, with each main person having such a verdict. Ultimately, 12.5 percent received a verdict mandating them to continue study or receive a harsher penalty such as strike-hard detention. The remaining main persons were either to be released, to “continue” in their state of release, or to first finish their minimum one-year period of “study” prior to release. Verdicts can also mandate changes to a person’s management area within a camp. In one instance, the persons’ verdict stated: “change to strict management” (调整为严管).

Figure 12. Source: Karakax List.

Of the 230 persons who were to be released, 48 explicitly received an appraisal to “continue” in this state (继续…). The China Cables VTIC management document specifically states that those who are released must remain under close supervision and ongoing evaluation. The Karakax List clearly testifies to the fact. Those who are released are being appraised in terms of their ongoing behavior, including participation in (mandatory) community activities. Examples include: “[person shows] stable thinking after graduating” (结业后思想稳定), “co-operates with all community activities” (配合社区各项活动), or “no other illegal behaviors of any kind were discovered during [this person’s] period of residence in our district” (在我辖区居住期间未发现参加任何违法违纪行为).

Release verdicts can take several different forms. Sometimes, they used theo word “graduate [from studies]” (结业). Frequently, they were more explicit in stating that a person was to be released into local (community or township) “management and control” (管控). This term is commonly used for the re-integration of former prisoners or detainees into society, and means that they are to remain under very close monitoring and supervision. The VTIC management document from the China Cables explicitly states that “graduates” are to be closely monitored for one full year, and that this involves “management and control”. (The exact same Chinese term (管控) is used). The local police authorities must “grasp [their] actual behavior in a timely manner; they must take their responsibility to subject [graduates] to ongoing educational assistance and management and control [管控] very seriously, even one person must not fall through the cracks, there must be full coverage.”[100]

“Graduation” or release into “management and control” can take different forms. The Karakax List mentions two instances in which camp detainees are directly transferred (业后移交) from their re-education camp to local regular vocational schools. Subsequent verdicts indicate that these “students” continue to be subjected to ongoing monitoring, with one verdict stating that “during the time at the school there were no law-violating behaviors” (学校期间无违法乱纪行为).[101]

Finally, “release” is often explicitly linked to subsequent employment. Of the 230 release verdicts, 50 explicitly mentioned release into employment (or permission to remain in ongoing employment). Of these 50, 38 specifically stated “industrial park employment” (园区就业). One particular verdict indicates a clear case of forced labor. This person is allowed to “work in the industrial park, but not permitted to graduate” (在园区就业但不同意结业).[102] The final verdict is that this person is now “trustworthy” and consequently permitted to continue to work at the park. This piece of information provides evidence confirms the author’s suspicion that many of those who supposedly “graduated” from VTICs were at least initially still kept in the same internment conditions. In any case, it is clear that the employment performance of VTIC “graduates” is subject to close monitoring and evaluation. The verdict of another case notes that the person “obeys the industrial park management” (服从园区管理), and is therefore allowed to “continue to work at the industrial park” (继续园区就业).[103]

Verdicts are subject to ongoing evaluations. The document shows 11 instances where a previous release verdict was reversed, often indicated by the explicit wording “disagree …” (不同意) and then a repetition of the past verdict phrase. Reversals can again revert, fluctuating in close correspondence with the authorities’ appraisals of the behavior of the interned person and their family. In one case with five data rows, the first verdict recommends “return home [into] local-level control” (回乡管控).[104] The verdict notes that the “actual behavior of [the person’s] family members is very good” (亲属…现实表现良好), and that they “eagerly co-operate with community work” (积极配合社区工作). The second data row explicitly reversed this first verdict with a “continued study (disagree with returning home into management and control)” (继续培训 (不同意回乡管控)). The brief verdict text only notes that the person’s “family relationships are complicated” (家庭关系复杂), a reason that is also found in other verdict reversal cases. The third verdict simply cites the original reasons for internment and leaves the “continued study” assessment unchanged. The fourth data row then again recommends release, citing the exact same reason from the fist data row: “actual behavior of [the person’s] family members is very good”. The fifth data row then again reverses the verdict to “continued study”. The main reason for this change is most likely the fact that this person was moved from a “regular management area” in the camp (third data row) to the strictest area, a fact that is highlighted in the verdict text. According to the VTIC management document from the China Cables, only persons who are in the regular area are permitted to graduate.

Verdicts can also become more damning as the evidence increases. One person, interned for being a “two-faced cadre”, was initially going to be released due to age (65 years) and various illnesses.[105] However, the next verdict produces evidence of his children’s behaviors which he, as a cadre, should have prevented. Two of his daughters veiled themselves between 2014 and 2015, and his son had eight instead of the legally permitted two children. One of his children was now accused of having participated in an illegal religious organization, and the verdict notes that this cadre has been hiding information without reporting it. The family is described as showing a strong anti-Han sentiment and possessing a dense religious atmosphere. The third verdict repeats these allegations and notes that this cadre is now in the camp’s “very strict” management area.

The document contains one instance where a person had to return to a camp.[106] The verdict text notes that “this person often gets into fights and does not cooperateive with the community’s inspection work” (本人经常打架斗殴,不配合社区清查工作).

Other verdict texts reveal some of the problems that the internment campaign has created. For example, one person is required to continue in the camp because he is said to be “holding a grudge over his brother’s paralysis, there is an anticipation [that he could take] revenge on society.” The document states that this brother participated in the June 26, 2013 incident in Hanerike Township (Hotan Prefecture). He must subsequently have spent time in a re-education facility, because it then says that he was “released from study on June 22, 2018 for physical health reasons (paralysis).” We can only speculate why and how the brother became paralyzed during internment.

Other verdicts reflect the fact that the state not only detains persons who are even slightly suspect, but also prefers to keep them interned when any such doubts remain. One person for example was simply sent to re-education because he had previously been in prison.[107] This person’s verdict states that they should remain in the camp because “the [cases or issues] that they are related too are comparatively complex”. In reality, the state harbors suspicions towards this person but apparently lacks any real evidence. Rather than acting on the principle of the presumption of innocence, the state prefers to presume guilt. The re-education system is the ideal place where it can “park” such people. Not only are the camps places where people are kept under complete control and hence pose no risk to society; they are also loci of ongoing transformation, meaning the longer a person stays there, the closer they become aligned (in theory) with CCP-sanctioned ways of thinking and acting.

Typically, intermittent or final verdicts that recommend ongoing study note that the person in question has not experienced sufficient changes in their thinking or attitudes, or, more rarely, that there is a lack of cooperation in some respect. Frequently, they cite the person’s past misdemeanors without noting anything positive in addition to them. One case exemplifies this aptly.[108] There, the initial verdict was “release”, noting that the person’s “thinking was greatly transformed“ (思想转变大). The second verdict, however, recommended “continued study”. The second verdict, however, recommended “continued study” and cited two facts to support that decision. On the one hand, the person’s “thought transformation” was now only deemed to be “average” (思想转变一般). On the other hand, their family members had not participated in the mandatory community flag raising events in a “timely manner”.

Besides the perceived behavior and thinking of the person in re-education, the ultimate verdict that they receive appears to be dependent on two other factors: a) the behavior of their family members, and b) the original reasons for their internment.

Family behavioral performance is explicitly named in 59.7 percent of all 311 cases, and in 67.0 percent of cases where the person is not already in a state of ongoing release. Family performance becomes much less relevant when a person was already released and is under ongoing monitoring. In those cases, verdicts typically only mention that person’s behavior back in the community (see examples mentioned above). Otherwise, family behavior is crucial. In a number of instances, improved family performance is the only reason given for a positive verdict. The reason for this is that the authorities are apparently wary to release persons back into family contexts where they might again fall into religious thought or behavior patterns. Both original reasons for internment and verdict texts very commonly mention the “religious atmosphere” of person’s families.

Generally, there appears to be a broad correlation between the nature of persons’ ultimate verdict as shown in the document, and the originally stated reason for their internment (Table 4). Cases where religion featured in the original internment explanation are twice as likely to result in a continued internment verdict as the total sample (25.3 versus 12.5 percent, for all 311 main persons). Similarly, 21.0 percent of those with a “link abroad” internment reason received a continued internment verdict, far higher than the average. However, when removing all those from this sub-sample who also had a religion-related internment reason, that share drops sharply to 6.3 percent.

Original internment reason(s) Continued internment verdict
Total sample 12.5%
“Untrustworthy” (all cases) 5.2%
“Untrustworthy” (no other reasons) 2.8%
“Untrustworthy” (and religion-related) 11.5%
Religion-related (all cases) 25.3%
Link abroad (all cases) 21.0%
Link abroad (only not religion-related cases) 6.3%
Link to detained persons (all cases) 16.7%

Table 4. Continued interned verdicts exclude verdicts that simply mandate completion of the minimum one-year internment period.

Overall, it seems evident that a religion-related original internment reason makes it far more likely for interned persons to be forced into ongoing re-education. This seems logical, given that the state’s primary concerns with persons who have been abroad or maintain some kind of connection abroad are related to religious practice. The government list of 26 “focus countries” squarely centers on nations with significant Muslim populations, and the state’s suspicions toward Uyghurs’ connections abroad are closely related to concerns over religious radicalization.

Similarly, there is an apparent correlation between age and verdict result. Among main persons aged 19 to 29 years, the share of those consigned to ongoing study (or strike-hard detention) stood at only 5.9 percent, versus 12.0 percent for those aged 30 to 44 years, and 22.1 percent for those aged 45 to 65 years. However, this correlation is more than likely superficial because persons in the higher age groups also much more commonly had religion-related reasons for internment.

In order to assess the effects of different variables on release or non-release, the author employed a logistic regression model. Release versus non-release was the binary dependent variable, and age (metric) as well as three of the binary coding categories for original internment reasons, “untrustworthy”, “religion-related”, and “link abroad” were the explanatory variables.[109] In a model without interaction effects between the explanatory variables, “religion-related” emerged as the only statistically significant explanatory variable for release versus non-release, although having a link abroad came close to the required 95% confidence interval level (Table 5). “Religion-related” again remained as the only statistically significant explanatory variable after taking into account level 2 interaction effects, although age came close to the required 95% confidence interval level (Table 6). At the same time, the relatively low goodness of fit values of the model indicate that much of the variation in regard to the verdict cannot be explained by any of these variables, and is to be attributed to other variables: in particular, the external and internal changes of the interned person and of their family members, as evaluated by the state.[110]

Source Value Standard error Wald Chi-Square Pr > Chi² Wald Lower bound (95%) Wald Upper bound (95%) Odds ratio
Intercept 2.76 0.83 11.02 0.00 1.13 4.39
Age -0.01 0.02 0.33 0.57 -0.05 0.03 0.99
Untrustworthy 0.74 0.55 1.77 0.18 -0.35 1.82 2.09
Religion-related .08 0.38 7.96 0.00 -1.83 -0.33 0.34
link abroad -0.68 0.39 3.01 0.08 -1.46 0.09 0.50

Table 5. Logistic regression of the relationships between age and reasons for internment on release verdicts (no interactions between explanatory variables). R²(McFadden) = 0.09, R²(Cox and Snell) = 0.06, and R²(Nagelkerke) = 0.12. The low goodness of fit values compared to table 6 indicate that interaction effects between explanatory variables are significant; also, that there are other, pertinent variables with a systematic effect: the perceived behavior of detainees and their families, along with the idiosyncrasies of each case.

Source Value Standard error Wald Chi-Square Pr > Chi² Wald Lower bound (95%) Wald Upper bound (95%) Odds ratio
Intercept 4.69 1.48 10.01 0.00 1.78 7.60
Age -0.06 0.03 3.21 0.07 -0.13 0.01 0.94
Untrustworthy 0.79 2.37 0.11 0.74 -3.87 5.44 2.20
Religion-related -4.05 1.78 5.18 0.02 -7.53 -0.56 0.02
link abroad -1.65 1.81 0.83 0.36 -5.19 1.90 0.19
Age*Untrustworthy 0.00 0.07 0.00 0.96 -0.15 0.14 1.00
Age*Religion-related 0.07 0.04 3.02 0.08 -0.01 0.15 1.07
Age*link abroad 0.02 0.04 0.21 0.65 -0.07 0.11 1.02
Untrustworthy*Religion-related -0.33 1.15 0.08 0.77 -2.58 1.91 0.72
Religion-related*link abroad 0.32 0.88 0.13 0.72 -1.41 2.05 1.37

Table 6. Logistic regression of the relationships between age and reasons for internment on release verdicts (level 2 interactions between explanatory variables). R²(McFadden) = 0.24, R²(Cox and Snell) = 0.17, and R²(Nagelkerke) = 0.31. The relatively low goodness of fit values indicate that there are other, pertinent variables with a systematic effect: the perceived behavior of detainees and their families, which are routinely mentioned, along with the idiosyncrasies of each case.

The insights gained from analyzing verdicts point to the essential ongoing role of village-based work teams, not only a role of gathering evidence that leads to internment. A spreadsheet from the Beshtoghraq ( 拜什托格拉克) village work team, a township in Lop County, Hotan Prefecture, designed to gather information about released persons and their family members, has a special column titled “actual behavior” (现实表现).[111] This particular term is found in nearly all verdict texts and can apply to the evaluation of either individual or family behavior. The spreadsheet then has work teams appraise these persons as being in one of four categories, from A to D.

Figure 13. Source: Source reference code: 62451.

A second file from the same region shows a version of this spreadsheet with actual data (Figure 14). It indicates that D is the best appraisal level, indicating that a person is permitted to be at home. Level C means that persons are supposed to be in VTICs, level B stands for VTICs or imprisonment, and level A, presumably the worst, is not found in the file. The “actual behavior” assessment only applies to those who are currently at home. In each instance below, this behavior was rated as “average” (一般).

Figure 14. Source: Source reference code: 62451. Only shows the last 26 of a total of 76 entries. Columns containing names and ID numbers were shortened in order to protect the identities of the persons shown.

Similarly, a government document from Hotan Prefecture from April 2019 notes that work teams are to assess the “actual behavior” (现实表现) of households, as does a village-based work team regulation from Karakax County from March 2016. This explains why Chen Quanguo declared the village-based work teams to be an essential long-term component of Xinjiang’s stability maintenance strategy.

5.0 Conclusions and Overall Significance

More than any other piece of publicly available evidence, the Karakax List presents us with an intimate picture of the inner workings of Beijing’s surreptitious battle for control over innermost human terrains of identity and allegiance. It dramatically reflects the extent to which the Chinese Communist Party has codified its intrinsic fear of religiosity and ethnic difference into a sophisticated set of internally consistent, quasi-scientific criteria for internment and release. This set of criteria constitutes the core of Beijing’s extrajudicial internment drive in Xinjiang.

Driven by the underlying principles of the presumption of guilt (rather than of innocence) and guilt by association, Beijing has managed to intern a targeted subset of the region’s Turkic minority populations in a way that holds entire minority populations hostage to their own behavioral performances. Through the ways in which release verdicts are assessed, whole family units are forced into a continuous mode of pre-emptive obedience. They are made to shun even the slightest observable forms of religiosity, knowing that it could lead to new internments, or prolong existing ones.

The policies implemented under Chen Quanguo have enabled the state to gain control over precisely those areas that Beijing has traditionally struggled with: control over religious practice; control over other markers of ethno-racial difference such as language, traditions, and dress; control over movement; and control over intergenerational transmission.

Can this work? Will Beijing’s re-education and social re-engineering drive in Xinjiang raise up a brainwashed, docile playbook version of minorities, a “model for the world to emulate” as touted by state media?

On the surface, the answer might be yes. If anything, the Karakax List amply demonstrates that anyone who does not conform to the state’s scheme of micromanaged social coercion has virtually no chance of escaping the system. However, by replacing trust with control, Beijing has arguably traded short-term gains for long-term jeopardies.

The first jeopardy pertains to mechanism of governance. While employing advanced technology at virtually every step of the process, the Karakax List shows that the whole scheme relies heavily on a labor-intensive micromanagement of minority communities. Given that oppression does not create mutual trust, Chen Quanguo’s words that the village-based work teams are a long-term component of Xinjiang’s stability maintenance should be taken at face value. That means that the state’s ability to maintain total control is predicated upon its ongoing ability to harness and finance significant numbers of loyal cadres who are willing to continually intrude into the private lives of the inferior citizenry. This mechanism itself is predicated upon unequal power relations, and hence perpetuates them by design.

This latter fact points to the second jeopardy. Beijing’s governance approach in Xinjiang is one of control and coercion rather than mutual trust. Presently, the state has maneuvered itself into a situation where it cannot but keep its minority underclass in a quasi-narcotic state of compliant passivity and numb acquiescence, with no opportunity to process the trauma of their captivity. Not even their own four walls is a safe place for venting anger, hurt, or pain. Many victims of re-education internment report symptoms of post-traumatic stress disorder. All this is hardly a recipe for long-term social stability. In the past, anger and frustration were vented through various means, including acts of violence. Will they now just remain bottled up?

Regardless of these two systematic risks, the long-term success of Beijing’s strategy in Xinjiang relies on a near-perfect substitution of intergenerational transmission. As parental generations are severely weakened and forced into submission, a new generation of young Uyghurs is being raised in educational institutions that are tightly controlled by the Party. This strategy, however, will not likely succeed. The main reason for this is that Uyghurs and other ethnic minority groups are still identified as such and therefore know that they are somehow different. This sense of difference is greatly exacerbated by the fact that the Han majority continues to treat them as second-class citizens. Research by the author and others into the identity of minorities who grow up in Han-dominated contexts shows that the experience of being different reinforces their original identity.[112] Given that the governance mechanisms that undergird China’s control over Xinjiang prescribe the perpetuation of such asymmetric ethnic relations, they represent a powerful force that could largely nullify the effects of intergenerational separation.

Beijing would need to ship minority infants to eastern Chinese communities and have them grow up as if they were Han and enjoyed completely equal rights. Assimilation would take place as bonds of mutual trust and equality flourish. However, the exact opposite is taking place. Sooner or later, the next generation of Uyghurs will actively try to find out who they truly are. The answer to that burning question will be found among the oppressed older generations.

At best, Beijing will succeed in splitting these ethnic groups into those who willingly integrate and enjoy the related benefits, and those who resist. However, there is already a generation of Uyghurs who pursued exactly such integrationist strategies and were still met with persecution and internment. The fact that Beijing did not even trust those who willingly opted for economic, cultural, and even political integration has sent one overarching message to the Uyghur communities: they will never be trusted, and they will never be accepted as equal. Their submission only temporarily allays the intrinsic fears of their masters.

With its most recent policies, Beijing has both exacerbated and cemented existing asymmetries to the point where equal relationships of mutual trust are extremely unlikely to develop. Consequently, the relationship between the state and the Turkic minorities in Xinjiang is both complex and difficult to predict in the long-term.

Adrian Zenz is a Senior Fellow in China Studies at the Victims of Communism Memorial Foundation, Washington, D.C. (non-resident), and supervises PhD students at the European School of Culture and Theology, Korntal, Germany. He specializes in China’s ethnic minority policy, minority education systems, public recruitment (especially teacher and police/security-related recruitment), public bid documentation, domestic security budgets and securitization practices in China’s Tibetan regions and Xinjiang. He has authored Tibetanness under Threat (Global Oriental, 2013) and co-edited Mapping Amdo: Dynamics of Change (Prague: Oriental Institute, 2017).

[1] The author would like to thank: his research assistants Cheryl Yu and Alexander Thilmany for their hard work, without which this report would not have been possible; Elise Anderson, senior program officer at the Uyghur Human Rights Project, for assistance with the analysis of transliterated Uyghur names and with the transliteration of Uyghur place names.

[2] Approximately 2,800 adults plus several hundred minors.

[3] Compare, page 10.

[4] Source reference code: 74128.

[5] See for further details. The sources from this file cache are referenced through source reference codes in order to protect ongoing research.


[7] Xinjiang 2018 statistical yearbook, table 3-7.

[8] or


[10] or

[11] and

[12] “表4-1至4-3.xlsx “ from or


[14] “表4-1至4-3.xlsx “ from or

[15] “表4-1至4-3.xlsx “ from or


[17] “表4-1至4-3.xlsx “ from or

[18] “表4-1至4-3.xlsx “ from or See also the business registration details of the Karakax County Bosidankule Aziz Bakur Restaurant (墨玉县博斯坦库勒艾则孜巴柯尔餐厅), which is located in district no. 12 (墨玉县博斯坦库勒十二区培训中心):

[19] or A substantial number of additional districts were added in 2016 and 2017.

[20] or

[21] Document “截至2017年政府性债务及置换债券情况说明.docx” from or

[22] “表4-1至4-3.xlsx “ from or

[23] 3墨玉县司法局部门单位项目预算支出绩效自评报告.docx from or

[24] Location: 37°13’42.55″N  79°44’14.44″E

[25] or, compare


[27] See

[28]  or

[29] or


[31] A detailed explanation of these three approaches can be found at or


[33] Compare related studies by Darren Byler,, and James Leibold,




[37] Compare also the statement “自“访惠聚”活动开展以来,特别是 2016 年以 来,财政厅工作组以强化基层组织建设为目标”, or

[38] or

[39] or

[40] or, compare or

[41] or


[43] or

[44] or

[45] Source reference code: 68305.


[47] or At times also used in a tripartite form: “remove the cover, eliminate the nest, dig behind the curtain” (揭盖子、端窝子、挖幕后).

[48] Source reference code: 10452.


[50] Source reference code: 77128.

[51] Source reference code: 76690.

[52]ÄêµØÇø±¾¼¶Ô¤Ë㵥λ²¿ÃžöË㹫¿ª²ÄÁÏ/2017Äê½Ì¿ÆÎIJ¿ÃžöË㹫¿ªµ¥Î»²ÄÁÏ(61)/µØί/н®Î¬Îá¶û×ÔÖÎÇøÖй²ºÍÌïµØί°ì¹«ÊÒ£¨±¾¼¶£©_ºÍÌïµØί°ì¹«ÊÒ2017Ä겿ÞöËã·ÖÎö±¨¸æ1.doc or

[53] or

[54] Source reference code: 50042. Compare or for use of the expression “切断宗教极端思想传承体系”.


[56] For a detailed study of the workings of the IJOP, see the related report by Human Rights Watch:


[58] See also or

[59] Source reference code: 24366.

[60] Source reference code: 34912.


[62] Source reference code: 32114.

[63] Numerous examples include,, or

[64] See e.g. or

[65] For example source reference code: 14156.

[66] Xinjiang Caikuai magazine, 2017 (issue no. 4), articles under the section “fanghuiju activity”, source:


[68] Source reference code: 83715.

[69] andÄêµØÇø±¾¼¶Ô¤Ë㵥λ²¿ÃžöË㹫¿ª²ÄÁÏ/2017Äê½Ì¿ÆÎIJ¿ÃžöË㹫¿ªµ¥Î»²ÄÁÏ(61)/µØί/н®Î¬Îá¶û×ÔÖÎÇøÖй²ºÍÌïµØί°ì¹«ÊÒ£¨±¾¼¶£©_ºÍÌïµØί°ì¹«ÊÒ2017Ä겿ÞöËã·ÖÎö±¨¸æ1.doc or

[70] and – 若羌县财政项目支出绩效自评报告(社区矫治康复日常管理项目).docx or

[71] Source reference code: 49105.

[72] ID 6532221992092*****





[77] ID number checksum verification was performed via

[78] There are far less sequential code combinations than 1,000. The author estimated them conservatively at 166 based on variation in the actual document.





[83] See and


[85] or, or

[86] See e.g. or

[87] ID no. 6532221966030*****.



[90] ID no. 6532221992030*****.


[92] or

[93] Source reference code: 19422.

[94] or

[95] or

[96] Numerous conversations of the author with visitors to Xinjiang, as well as persons in China with close contacts in southern Xinjiang.

[97] ID no. 6532221988040*****.

[98] Source reference code: 73718.

[99] or; or

[100] Original section in the document: 21. 跟踪帮教。加强结业学员跟踪帮教工作,由所在地基层组织负责跟踪帮教,派出所、司法所纳入列管对象,一年内不得脱离视线,及时掌握现实表现。要严格落实帮教和管控责任,做到一人不漏,全面覆盖。

[101] ID no. 6532221994080*****

[102] ID no. 6532221984010*****

[103] ID no. 6532221991011*****

[104] ID no. 6532221975051*****

[105] ID no. 6532221954082*****

[106] ID no. 6532221996013*****

[107] ID no. 6532221983041*****

[108] ID no. 6532221978080*****

[109] Minimum sample size requirement was computed as N = 10 k / p = 10 * 4 / (62/310) = 200. The actual sample size was 310. A regression of more coded internment reasons with fewer occurrences for main persons would require a sample size that exceeds the available figure. The resulting model showed a -2 Log(Likelihood) (Chi² 56) p < 0.0001.

[110] R²(McFadden) = 0.24, R²(Nagelkerke) = 0.31.

[111] Source reference code: 62451.

[112] See e.g., and Grose, Timothy, 2015. ” (Re)Embracing Islam in Neidi: the ‘Xinjiang Class’ and the dynamics of Uyghur ethno-national identity”, Journal of Contemporary China, Vol. 24, No. 91, 101–118,