BEIJING (Reuters) – A Reuters review of government documents shows that dozens of Chinese companies have made software that uses artificial intelligence to filter data collected on the population, amid heavy demand from authorities seeking to update monitoring tools.
According to more than 50 publicly available documents examined by Reuters, dozens of entities in China over the past four years have purchased such “one person, one file” programs. Technology improves on existing software, which simply collects data but leaves it for people to organize.
“The system has the ability to learn independently and can improve the accuracy of file creation while increasing the amount of data. (Faces) are partially obscured, disguised or wearing glasses, and low-resolution images can also be archived relatively accurately,” according to a tender published in July by the Public Security Administration. In Henan, the third most populous province in China. Read more
Register now to get free unlimited access to Reuters.com
Henan’s public security department did not respond to requests for comment on the system and its uses.
The new program improves Beijing’s current approach to surveillance. Although current systems in China can collect data on individuals, law enforcement and other users are left to regulating them.
Another limitation of current monitoring programs is their inability to link an individual’s personal details to a real-time location except for security checkpoints such as those at airports, according to Jeffrey Ding, a postdoctoral fellow at the Stanford Center on International Security and Cooperation.
One person, one file “is a way of sorting information that makes it easier to trace people,” said Marika Olberg, a senior fellow at the German Marshall Fund in Berlin.
China’s Public Security Administration, which oversees provincial police authorities, did not respond to a request for comment on one person, one file and their surveillance uses. Besides police units, CPC bodies responsible for political and legal affairs opened 10 bids. China’s Central Political and Legal Affairs Commission declined to comment.
The tenders examined by Reuters are just a small part of these efforts by Chinese police units and party bodies to upgrade surveillance networks by harnessing the power of big data and artificial intelligence, according to three industry experts interviewed for this story.
According to government documents, some users of the software, such as schools, wanted to monitor unfamiliar faces outside their compounds.
The majority, such as police units in Ngawa Prefecture in southwestern Sichuan Province, which is mainly inhabited by Tibetans, ordered this for more explicit security purposes. Nagawa’s tender describes the software as aiming to “maintain political security, social stability and peace among people”.
The Public Security Department of Nagawa did not respond to requests for comment.
Beijing says its monitoring is critical to fighting crime and has been central to its efforts to combat the spread of COVID-19. Human rights activists such as Human Rights Watch say the state is building a privacy-invasive surveillance state that unfairly targets certain groups, such as the Uyghur Muslim minority.
A Reuters review shows that local authorities across the country, including densely populated areas in Beijing and underdeveloped provinces such as Gansu, have opened at least 50 tenders in the four years since the first patent application, 32 of which were opened for tender in 2021. Now 22 tech companies, including Sensetime, Huawei, Megvii, Cloudwalk, Dahua and Baidu’s cloud division, have such software, according to a Reuters review.
Sensetime declined to comment. Megvii, Cloudwalk, Dahua, and Baidu’s cloud division did not respond to requests for comment.
Huawei said in a statement that one of the partners has developed the One-Person, One-File application in its smart city platform. The company declined to comment on patent applications.
“Huawei does not develop or sell applications that target any specific group of people,” the company said.
The documents reviewed by Reuters span 22 of China’s 31 major administrative departments, all levels of provincial government, from regional public security departments to party offices in a single district.
The new systems aim to make sense of the vast amount of data these entities collect, using complex algorithms and machine learning to create customized files for individuals, according to government tenders. The files update themselves automatically as the program sorts the data.
A wide range of challenges can complicate implementation. Three experts in artificial intelligence and monitoring told Reuters that bureaucracy and even cost could create a fragmented and disjointed nationwide network.
Reuters found ads for successful bids for more than half of the 50 procurement documents analyzed, with values ranging from a few million yuan to nearly 200 million yuan.
China covered its cities with surveillance cameras in the 2015-2020 campaign it called “sharp eyes” and is striving to do the same across rural areas. The development and approval of the One Person, One File program began around the same time.
The first mention of a single person, said Olberg, was a single file from 2016, in a 200-page surveillance feasibility study by Xinjiang’s Shaoan County, for a computer system that could “automatically identify and investigate key people involved in terrorism and (It threatens social stability). An official of Shauan County declined to comment.
In 2016, the head of China’s internal security at the time, Meng Jianzhu, wrote in a state-run magazine that big data was the key to finding patterns and trends in crime. Two years later, the system was referenced in a letter to industry executives by Li Ziqing, director of the Biometrics Research and Security Technology Center at the state-run Chinese Academy of Sciences. Lee was also chief scientist at AuthenMetric, a Beijing-based facial recognition company. Neither the think tank nor AuthenMetric responded to requests for comment.
“The ultimate core technology for big data security (application) is one person, one file,” Li said in a 2018 speech at the Artificial Intelligence Forum in Shenzhen, according to the transcript of the speech published by local media and shared on AuthenMetric’s WeChat website. public account.
The party’s Political and Legal Affairs Committee, which Meng led in 2016, declined to comment. Meng could not be reached for comment. I did not respond to a request for comment.
The industry developed rapidly. By 2021, Huawei, Sensetime, and 26 other Chinese technology companies have filed patent applications with the World Intellectual Property Organization for file archiving and image aggregation algorithms.
Huawei’s 2021 patent application for “People Database Segmentation Method and Device” mentioning one person, one of the files said that “As smart cameras become more and more popular in the future, the number of face photos taken in a city will increase to trillions annually.”
The 50 bids analyzed by Reuters provide varying amounts of detail about how the software will be used.
Some have mentioned “one person, one file” as a single entry in the list of items needed for monitoring systems. Others gave a detailed description.
Nine of the tenders indicated that the software would be used with facial recognition technology that, according to the specific documents, could determine if a bystander was Uyghur, connect to police early warning systems and create archives of Uyghur faces.
For example, a bid published in February 2020 by the party apparatus in charge of a district in the southeastern province of Hainan Island, sought a database of Uyghur and Tibetan residents to facilitate “finding the information of people involved in terrorism.”
Hainan authorities did not respond to a request for comment.
More than a dozen tenders refer to the need to combat terrorism and “maintain stability,” a general term that human rights activists say is often used to suppress dissent.
At least four of the tenders said the software should be able to pull information from individuals’ social media accounts. Half of the bidders said the software would be used to collect and analyze personal details such as relatives, social circles, vehicle records, marital status and shopping habits.
Register now to get free unlimited access to Reuters.com
(Reporting by Eduardo Baptista). Editing by Jerry Doyle and Brenda Goh
Our Standards: Thomson Reuters Trust Principles.