Chinese officials everywhere are introducing high-tech surveillance of ordinary citizens, but they prefer to keep their affairs in silence. Deployed in the pilot version of the AI system Zero Trust found thousands of corrupt officials – and their colleagues did not like it.
Great material on the short-term take-off and decline of the Zero Trust algorithm in China is published by the South China Morning Post. The system of identifying discrepancies in the sea of information on the income and expenses of officials quickly became so effective that it was turned off.
The first version of Zero Trust was launched in 2012 – just when Chairman Xi Jinping took over the top post, replacing Hu Jintao. The scope of the system was modest – only 30 cities and districts of China. It covered less than 1% of all civil servants. And despite this, the algorithm caught almost 9,000 officials by the hand.
Zero Trust compares multiple databases with revenues and expenditures and government tenders. Initially, the algorithm was trained by employees of the Chinese Communist Party’s internal investigations department. They manually marked data on typical inconsistencies in various statements that indicate the sins of officials. Then the algorithm was fed more and more information. Now it has access to about 150 databases at the federal and regional levels.
If necessary, Zero Trust is even checked with satellite photos in order to find out if the road specified in the documentation to a remote village was built or the money settled in the pockets of officials.
The system perfectly identifies typical offenses: illegal transfer of ownership, inconsistency of infrastructure with the application, fraud with the land and the destruction of private houses, participation in the tender of the company by relatives of a responsible official, or an unexpected change in the bank account of the civil servant. Even if the data in one document is forged, inconsistencies in the conjugate statements will quickly indicate this.
As in the case of other modern AI, Zero Trust is often difficult to “explain” why an official received a black mark. Therefore, operators work side by side with it. And the last word is always for a man.
In other areas of life, the authorities increasingly rely on the help of surveillance cameras and smart algorithms: the social rating system covers more and more citizens and private companies, and in Guizhou data on the movement of each police officer instantly enters the cloud. It turns out that it is much more important that Big Brother not see the movement of money.
Those involved in the work of the Zero Trust claim that they used the AI diplomatically and gently. For example, they describe the result of their work not as a prison sentence or a fine for a corrupt official, but as his “warning in the early stages” of such activity. The system was deployed on only 1% of the area, and it was mostly about the poorest provinces away from the political “centers of power”.
Since 2012, the system has caught the hand of 8,721 officials who were accused of abuse of power, misappropriation of funds or abuse of authority.
Only a small part of them was sent to prison on orders from the AI. The majority, SMTP says, even kept their jobs, having received only reprimands.
However, even this “soft” version for the officials was not to their liking. In most of the provinces where they were tested, Zero Trust is now disconnected. Jiang Yi, a spokesman for the Commission for the Inspection of Discipline from Hunan Province, said that his department continues to use ZT, but literally faces pressure every day. He is one of the few SMTP commentators whose name is named in a publication. His anonymous colleague says that none of the officials at first do not even want to hear about submitting data from their department voluntarily, and it requires “a little pressure” to get them.
More and more officials are outraged that the AI does not need the signature of the higher management or the judge’s verdict to access their personal data.
As a result, ZT now continues to fight corruption in only a few districts. And, apparently, it will never work across the country – as the creators of the algorithm once hoped.