China uses AI software to improve its surveillance capabilities

BEIJING – Dozens of Chinese companies have developed software that uses artificial intelligence (AI) to sort residents’ collected data, amid high demand from authorities to upgrade their surveillance equipment, Reuters reviews official documents.

According to more than 50 publicly available documents examined by Reuters, dozens of Chinese companies have purchased such software over the past four years, known as “one person, one file”. The technology improves existing software, which not only collects data but also leaves it to people to organize.

“The system has the ability to learn independently and can optimize file creation accuracy as data volume increases. [Faces that are] Partially blocked, masked or spectacled, and low-resolution portraits can also be comparatively accurately preserved, ”according to a tender published in July by the Department of Public Security in Henan, China’s third-largest province by population.

Henan’s Department of Public Security did not respond to requests for comment on the system or its use.

The new software improves Beijing’s current surveillance system. Although China’s existing systems can collect information from individuals, law enforcement and other users are left to organize it.

According to Jeffrey Ding, a postdoctoral fellow at Stanford’s Center for International Security and Co-operation, another limitation of current surveillance software is the inability to link a person’s personal details to a real-time location without security checkpoints such as airports.

One person, a file is “a way of sorting information that makes it easier for people to track,” said Ohlberg, a Berlin-based senior fellow at the German Marshall Fund.

China’s Department of Public Security, which oversees regional police authorities, did not respond to requests for comment on a person, a file and its surveillance usage. In addition to the police unit, 10 bids were opened by the Chinese Communist Party organizations responsible for political and legal matters. China’s Central Political and Legal Commission declined to comment.

The tenders tested by Reuters represent a part of such efforts by Chinese police units and party agencies to upgrade surveillance networks by tapping into the power of Big Data and AI, according to three industry experts interviewed for the story.

According to official documents, some software users, such as schools, wanted to monitor unfamiliar faces outside of their compounds.

The majority, such as the police unit in Nagawa Prefecture in southwestern Sichuan Province, mainly by Tibetans, ordered it for more explicit security purposes. Engawa described the tender software as “for political security, social stability and maintaining peace among the people”.

Ngawa’s public safety department did not respond to a request for comment.

Beijing says its monitoring is crucial to the fight against crime and the key to its efforts to combat the spread of coronavirus disease 2019 (COVID-19). Human rights activists, such as Human Rights Watch, say the country is creating a surveillance state that violates privacy and unfairly targets certain groups, such as the Uighur Muslim minority.

Reuters reviews show that local authorities across the country, including Beijing’s high-populated districts and underdeveloped provinces like Gansu, have opened at least 50 tenders in the four years since the first patent application, 32 of which were open for bidding in 2021. Twenty-two technology companies now offer such software, according to a Reuters review.

Senstime declined to comment. The cloud division of Megvii, Cloudwalk, Dahua, and Baidu did not respond to a request for comment.

Huawei said in a statement that a partner has created a one-person, one-file application on its Smart City platform. The company declined to comment on the patent application.

“Huawei does not create or sell applications that target a specific group of people,” the company said.

Documents Reuters reviewed 22 of China’s 31 main administrative departments, and all levels of provincial government, from the regional public security department to the party office for a single neighborhood.

According to official tenders, the new systems aim to generate huge amounts of data collected by such companies using sophisticated algorithms and machine learning to create customized files for individuals. Files are updated automatically as the software sorts the data.

However, broader challenges can complicate implementation. Bureaucracy and even costs could create a fragmented and isolated nationwide network, three AI and surveillance experts told Reuters.

Reuters has announced bids for more than half of the 50 procurement documents analyzed with values ​​ranging from a few million yuan to 200 million yuan.

System upgrade

China has blanketed its cities with surveillance cameras in a 2015-2020 campaign that has been described as “sharp-eyed” and is trying to do the same in rural areas. The development and adoption of “one person, one file” software began at the same time.

Ms Ohlberg, the researcher, said she first mentioned a person, a file in a 2016, 200-page surveillance feasibility study in Xinjiang’s Shawan County, to obtain a computer system that could “automatically detect and investigate terrorism and (social) instability”. An official from Shawan County declined to comment.

In 2016, the then-head of China’s internal security, Meng Jianzhu, wrote in a state journal that there was a lot of information to find out the types and trends of crime. Two years later, the system was mentioned in a speech by industry executives given by Li Zhiking, then director of the research center for biometrics and security technology at the state-run Chinese Academy of Sciences. Li was also the chief scientist at Athenometric, a Beijing-based facial recognition company. Neither the research center nor the authoritative body responded to the request for comment.

“Big data is the ultimate core technology [application to] Security is one person, one file, ”Mr Lee said in a 2018 speech at an AI forum in Shenzhen, according to a transcript of a speech published by local media and shared on Athenometric’s WeChat public account.

The party’s political and legal affairs commission, which Mr Meng headed in 2016, declined to comment. Mr Meng could not be reached for comment. Mr Lee did not respond to a request for comment.

The industry is developing rapidly. By 2021, Huawei, Senstime and 26 other Chinese technology companies have filed patent applications with the World Intellectual Property Organization for file archiving and image clustering algorithms.

A 2021 Huawei patent application is a “Personal Database Partitioning System and Device” that specifies an individual, a file that states that “as smart cameras become more popular in the future, the number of facial images captured in a city will be trillions each year.”

Safe city

The 50 Reuters-analyzed tenders give a wide range of details on how the software will be used.

Some have listed “one person, one file” as a single entry in the list of items needed for surveillance. Others have given detailed descriptions.

Nine tenders indicated that the software would be used in conjunction with facial recognition technology that, in certain documents, could identify a pedestrian as a Uyghur, connect to an early warning system for the police, and create an Uyghur facial archive.

For example, a tender published in February 2020 by a party organ responsible for an area in the southeastern island province of Hainan sought a database of Uighur and Tibetan residents to facilitate “finding information on individuals involved in terrorism.”

Hainan authorities did not respond to a request for comment.

More than a dozen tenders have mentioned the need to fight terrorism and “maintain stability”, a catch-all term often used by human rights activists to suppress dissent.

At least four tenders said the software should be able to extract information from an individual’s social media account. Half of the tenders said the software would be used to compile and analyze personal details such as relatives, social circles, car records, marital status and shopping habits. – Eduardo Baptista / Reuters