In China, countering corruption with an AI-based program has shown some success, with many local government officials pushing for its closure. The system, called Zero Trust, belongs to the Chinese Academy of Sciences. It is based on AI, and its main purpose is to suppress corruption among employees of the state apparatus. The program monitored and evaluated the actions of officials, and in some cases could interfere with their work.
At the direction of the Chinese authorities, the Zero Trust is closing. Civil servants who advocated the ban on the project cited as the reason that the AI system does not have the right to access classified information and that at the time the program is running, government officials are under great pressure.
The project started in 2012. Initially, large-scale use of the system throughout China was not planned, its coverage was only a few counties and cities. Despite the small territorial coverage (about 1% of the country), corruption in China in some cases was revealed using the latest IT technologies. During the period of its work, the AI was able to identify more than 8,700 people convicted of illegal actions. Some of them were punished, the other got off with warnings. In turn, the creators of the project emphasize that their development is not intended to punish, but is aimed at identifying the early stages of illegal actions and returning government officials to the “right path”.
The program, with the help of which the fight against corruption was carried out, worked with several dozen central and regional databases. As a result, Zero Trust was able to construct complex structures of social interactions and, on their basis, analyze the behavior of civil servants. If the employee was able to falsify any information, the AI compared the data from different sources, finding discrepancies.
Zero Trust has shown particular efficiency in identifying suspicious transactions involving the transfer of property, construction of infrastructure, demolition of housing and the acquisition of land. In certain cases, the program, implementing anti-corruption, could create a request for obtaining satellite images, for example, to check the fact of the construction of an object or a road in a certain settlement. The system also identified cases of employees' bank accounts and large purchases.
The program contains an algorithm that evaluates how fashionable a certain action can be attributed to illegal. One of the features of Zero Trust is that the system is able to detect corrupt officials, but cannot provide a good explanation for its findings. Although in many cases the algorithms of the program turned out to be correct. After identifying the intruder, the specialists monitoring the operation of the system could directly contact the civil servant with a recommendation not to make further mistakes. The creators of the project claim that the final decision remains with the official, while the program does not send the results of its investigation anywhere.
The Topic of Article: China shut down AI computer program that was too good at detecting corruption.