The Senate wants to create a European controller of the reliability of facial recognition algorithms

Senators Marc-Philippe Dobress (Union for a Popular Movement), Arnaud de Belene (Centrist Union) and Jérôme Durin (Group of Socialists, Ecologists and Republicans) presented on May 11 the results of a report on the use of biometric recognition in public space.

Spectrum of Mass Surveillance

“It is necessary to build a collective response to the use of biometric recognition technologies in order to keep up with industrial developments in the coming years,” they note in advance. To solve this problem, they made 30 proposals to “eliminate the risk of surveillance of society.”

The uses of facial recognition are potentially limitless. But at present, French law allows only a few of them in very specific contexts: the photo reconciliation device used in the Processing of Criminal Records (TAJ) and the Parafe border crossing system. Several experiments have also been carried out – by the City of Nice or Aéroports de Paris.

The authors of the report are still surprised that today there is no general regulation governing face recognition. Certain provisions are provided for in the General Data Protection Regulation (GDPR) when it concerns the protection of biometric data. Otherwise, they are specific texts that allow recourse to recognition in a very specific case, as in the case of TAJ.

Four prohibitions in the use of facial recognition

Faced with this observation, they first propose the development of a structure incorporating “red lines” to eliminate the risk of mass surveillance. Thus, four prohibitions must be established: the prohibition of social rating, the categorization of persons by ethnic origin, gender or sexual orientation (except in the context of scientific research), emotion analysis (except for medical and research purposes), and real-time mode. remote biometric monitoring in public places (with the exception of security forces in very limited cases).

To identify use cases in which this technology could be relevant, the report recommends enacting a law on experimentation for a period of three years. Thus, the government and parliament will be obliged to reassess the needs and rebuild the system in accordance with the results obtained. For their part, the French should be able to take advantage of “clear information about biometric recognition methods, the expected benefits from them and the risks associated with them.”

In addition, the use of recognition must be authorized a priori by the magistrate or prefect in the case of internal security forces, and by the National Commission for Computing and Liberties (Cnil) in the case of a private institution.

I prefer algorithms developed in Europe

For speakers, technological sovereignty should be at the heart of thinking about face recognition. Indeed, “the use of algorithms developed in Europe based on traceable data hosted on our soil, for example, is much preferable to the use of foreign algorithms, of which we sometimes know nothing about the conditions of creation and training.”

To strengthen sovereignty, the senators recommend that a European body be given a “mission to assess the reliability of biometric recognition algorithms and certify that they are free from bias.” The European Union appears to support this system as it proposes a similar one in its future rules on artificial intelligence, part of which concerns facial recognition. The only problem is that companies will have to agree to open the hood of their technology to make sure it’s reliable. This creates problems related to intellectual property, which is especially important in this sector.

Back to top button