The publication had been expected for several days and on Tuesday 4 February, the European Commission published its “guidelines” document on prohibited artificial intelligence practices, directly linked to Article 5 of the regulation on artificial intelligence (AI Act).
These guidelines are intended to increase legal clarity and provide an overview of the European Commission's interpretation of the prohibitions listed in the AI Act, “with a view to ensuring their consistent, effective and uniform application”.
More clearly, these must be accessible to companies that do not have access to a legal expert on the issue, and they must illustrate concrete cases where AI use is prohibited.
Most of them relate to the use of biometric information and respect for fundamental rights. They are superimposed on certain prohibitions under the Data Protection Regulation, such as Article 22 on “Automated individual decision making and profiling”.
These guidelines set out four main prohibitive categories:
- subliminal, purposefully manipulative or deceptive techniques” that influence people without their knowledge and encourage them to make decisions that may cause them “significant physical, psychological, financial or social harm”;
- the use of AI for “social scoring”, particularly the use of “unrelated data” to profile and classify an individual, or “profiling” on a larger scale to anticipate or predict an individual’s level of criminality based on their biometric information;
- the use of video surveillance images to create “facial recognition databases” or to monitor the psychological state of an employee, as well as their use in public places by the authorities for the purposes of maintaining public order;
- the use of biometric data to “assign people to categories” (political, sexual).
While these bans have been applicable since Sunday 2 February, the chapters on governance, enforcement and sanctions will not be applicable until 2 August. Member States will have until that date to designate the national market surveillance authority who will be responsible for the checks.
The European Commission reserves the right to update these guidelines, if necessary, according to feedback from the sector and technological developments in AI.
The guidelines on the definition of AI systems are still awaited.
See the document: https://aeur.eu/f/fct (Original version in French by Isalia Stieffatre)