China: AI Revolutionizes Its Surveillance Network

China has embarked on a massive modernization of its surveillance network by integrating artificial intelligence technologies into millions of existing cameras. This initiative aims to transform an aging system into a powerful tool capable of automatically analyzing behaviors and conducting video searches based on text queries.

This transformation is part of a 2024 directive issued after a series of violent attacks and marks a shift towards large-scale behavioral surveillance. According to human rights experts, this approach could significantly expand the scope of surveillance in China. Anthropic, an organization specializing in artificial intelligence, has warned of a possible intensification of this surveillance by 2028.

Local Chinese authorities have begun equipping old cameras with computer vision and language processing models. The Financial Times, after reviewing several procurement documents and interviewing industry players, reports that the police now have enhanced capabilities to automatically analyze behaviors and anticipate potential disturbances.

Since the mid-2010s, China has been using facial recognition and other computer vision technologies. However, the previous system, based on outdated hardware, was primarily reactive and sent data to centralized processing centers. Minxin Pei from Claremont McKenna College pointed out that this system was ineffective at predicting the intentions of individuals not listed on watch lists.

Companies like Hikvision and Huawei are now providing cameras equipped with advanced computer vision and language processing capabilities. These systems can detect abnormal behaviors, such as erratic driving or the formation of crowds, and generate automatic alerts. The latest generation of Hikvision cameras also allows agents to search for footage using text queries, thus simplifying the surveillance process.

A Hikvision official explained that the police no longer need to manually review video footage. The system can be powered by a simple text query to identify relevant clips, digitizing tasks that were previously manual.

Resource Optimization and Targeted Upgrades

A procurement document from Yaodu in Sichuan mentions the purchase of 175 HD cameras with intelligent video analysis. Additionally, a tender in Datong includes Hikvision cameras capable of determining individuals' gender, posture, and clothing. The initial implementations of these technologies focus on densely populated urban areas and the surroundings of military and government buildings. Some agencies are retaining their existing cameras but replacing intermediate servers with AI-equipped PCs to process videos locally, thereby reducing cloud-related costs.

These upgrades follow a 2024 directive from Public Security Minister Wang Xiaohong, issued after violent attacks attributed to a mental health crisis exacerbated by pandemic lockdowns and a stagnant economy. Minxin Pei noted that these incidents have revealed the limitations of the current surveillance system. Maya Wang from Human Rights Watch warned that the underlying philosophy of the Chinese surveillance system is becoming increasingly expansive. AI and computer vision provide authorities with unprecedented capabilities to monitor behaviors on a large scale.

In a recent policy document, Anthropic warned that if the technological gap continues to narrow, China could not only catch up by 2028 but also intensify AI-driven surveillance and repression.