Governments and companies are increasingly using artificial intelligence (AI), whether to allow access to applications or locations through biometric authentication or to perform automated medical diagnoses.
Faced with this situation, the National Consultative Commission on Human Rights (CNCDH) issued an opinion on April 7, 2022 on the impact of artificial intelligence on fundamental rights.
The need for a legal framework
Like the CNCDH, many international bodies have warned about the impact of AI on fundamental rights, particularly on respect for human dignity, privacy and data protection, equality and non-discrimination, access to justice and access to social rights.Currently, there is no legal framework, either at the national or international level, to prevent these risks.
In April 2021, the European Commission presented a proposal for a regulation establishing harmonized rules on artificial intelligence. This proposal still needs to be adopted by the European Parliament and the European Council. Work is also being carried out within the Council of Europe.
The CNCDH issues about twenty recommendations for the establishment of a binding legal framework, which is able to guarantee the respect of fundamental rights.
Drawing red lines for AI uses
"Certain uses of AI are too serious an infringement of fundamental rights to be allowed. It is up to the public authorities to prohibit their implementation".The CNCDH's opinion is consistent with the recent proposal for an EU regulation on AI, which lists uses that are "prohibited because they are contrary to the values of the Union.
The proposed EU AI regulation calls for a ban on:
- systems that rely on subliminal components that people cannot perceive, or that exploit the vulnerabilities of children and people who are vulnerable because of their age or physical or mental disabilities, and that, by altering their behavior, can cause them physical or psychological harm;
- AI systems allowing social rating of natural persons, based on their behavior or personal characteristics, by or on behalf of public authorities, for purposes of prejudicial or unfavorable treatment of certain natural persons or groups of persons;
- remote "real-time" biometric identification of individuals in publicly accessible areas for law enforcement purposes.
Prohibit certain uses
In its opinion, the CNCDH specifies and completes the list of prohibitions set out in the proposed European regulation.The CNCDH recommends " the prohibition of :
- The implementation of any type of social rating by administrations or public or private companies;
- Remote biometric identification of persons in public spaces and places accessible to the public;
- The use of emotion recognition technologies.
Framing the uses
"From design to end use of AI systems, the impact on fundamental rights must be regularly assessed.Public or private users must conduct impact assessments that include a human rights perspective, and that identify the risks of human rights violations - in particular discriminatory effects - as well as the mitigation measures put in place to address the risks involved.These studies should be complemented by stakeholder consultation, including those affected by the AI system."Guarantee a human intervention
"In order for AI to remain under human control, the CNCDH calls for the reintroduction of the human in the automated decision process: control of the result by the user and the right to parameterize the algorithm."Inform and train
"Any person affected by an AI system, whether the user is public or private, must be informed of the intervention of AI in the decision-making process concerning him or her and, in an intelligible way, of the way the algorithm works.As AI systems affect the whole population, the CNCDH considers that it is fundamental to develop information and training tools accessible to all, and to include the technical, political and societal issues of AI in citizenship education programs."Références :
