Artificial intelligence systems have taken a good place in the control, investigation and sanction activities carried out by law enforcement and public administrations, according to the Council of State.
In a recent study on public AI, written at the request of the government, the highest administrative court declares that AI systems offer “significant potential for the detection of infractions, breaches and threats (… ), the conduct of investigations as well as the organization of the preventive and repressive response”.
If the Council of State measures the risk of using this technology without safeguards, it notes on the other hand that the use of AI systems can “restore (r) a real efficiency of the control authorities”.
The hunt for fraud, the quest for ROI
Faced with the scale of the data to be processed, the General Directorate of Public Finances (DGFIP) has been talked about very recently with its “Innovative Land” project. It aims to detect undeclared buildings or facilities (such as swimming pools).
The DGFIP has been working on this fraud detection technique since 2017. The swimming pool detection system has already raised 10 million euros and will be rolled out to the whole of France from September.
When it comes to detecting fraud, the return on investment of technology is quite obvious, which partly explains its enthusiasm, notes the Council of State. Among the other “advanced” uses in the economic sphere, the report mentions the “Tracfin” intelligence service, piloted by Bercy, to fight against clandestine financial circuits, money laundering and the financing of terrorism.
Lots of expectations
Some public actors use open access data to report abuses and violations. This is particularly the case with the Ministry of Agriculture, which identifies establishments “at potential health risk” by automatically reviewing the opinions left by Internet users on the Internet. The Council of State also mentions the case of competition authorities, which plan to use AI systems to “detect and qualify agreements by algorithmic collusion, which can take the most diverse forms, from automated control of prices practiced within an organized system to the supply by the same service provider of the same algorithm to competitors allowing price alignment”.
But the most controversial use of AI, and also the most widespread, remains “the prevention of public order disturbances and the identification of suspects in public space by biometric recognition”, says the Council of state. We are talking here about tools allowing in particular to identify suspicious behavior by the exploitation of data flows, the detection of terrorist threats or the dissemination of false information by or on behalf of foreign powers with the Viginum service.
Although widespread, these AI systems are still inefficient, however, believes the Council of State. The authority pleads for a national strategy for public AI to be adopted, with “safeguards” commensurate with the potential of the technology.