On 21.4.2021, the European Commission published the proposal for a Regulation on Artificial Intelligence (the “Artificial Intelligence Act”).

Artificial Intelligence (AI) is the ability of a machine to display human-like capabilities such as learning and planning. AI enables technical systems to perceive their environment, deal with what they perceive, solve problems and act to achieve a specific goal.

Artificial Intelligence Act (AIA) defines Al systems as “software” that encompasses machine learning, expert and logic systems or statistical approaches and generates outputs such as content, predictions, or decisions that “influence” the environments it interacts with.

Under the AIA, a uniform legal framework for the use of AI systems is introduced, so that the safety and performance requirements for their introduction in the EU market, are fulfilled. To serve that purpose, AIA includes provisions which:

– classify the use of Al systems, depending on the potential risk (“unacceptable risk”, “high risk”, “minimal or no risk”) they pose to individuals and society. In particular, AI systems that are a threat to the safety, livelihoods and human rights, such as the ones used for the classification of people based on their social behavior, are considered as posing “unacceptable risk” and will be banned. “High-risk” AI systems, such as the ones used for the operation of critical infrastructure, for instance the management and operation of road traffic will be subject to additional obligations, which include a conformity assessment prior to those systems being marketed or put into service. Finally, AI systems that pose “minimal or no risk’’ for citizens’ rights or safety, such as AI-enabled video games, will be permitted with no further restrictions.

– establish governance systems for AI systems. At an EU level, the AIA establishes an European Artificial Intelligence Board, composed of representatives from the Member States and the Commission. At a national level, the AIA stipulates that Member States will choose either to appoint an existing or a new national authority which shall supervise the application and implementation of the AIA.

– encourage national competent authorities to set up AI regulatory sandboxes, so that innovative AI technologies can be tested in a controlled environment for a limited time, on the basis of a testing plan agreed with such authorities.

In general, the AIA – in the absence of prior AI regulatory framework- aims to address numerous aspects linked to the use of AI systems. It adopts a risk-based approach for their use, introduces new supervisory bodies and encourages innovation in AI technology.

However, there are still certain issues that require further clarification or consideration, such as:

a) the lack of definition for “high-risk” AI systems: the proposed AIA simply outlines the criteria to be used to determine whether an AI system should be classified as “high risk” and does not provide for any specific guidelines for their identification.

b) the unclear use of biometric identification systems: Under the AIA, biometric identification systems -through which an image of a person can be captured and compared to an existing database with the aim of identifying who the person in the image is -, are classified as posing “high risk”, whereas their “real time” use in publicly accessible spaces for law enforcement purposes is being prohibited with only few exceptions applying. Subsequently, the fact that the AIA does not expand the scope of the said prohibition to all forms biometric AI systems leaves substantial security gaps for the use of such systems by public authorities and businesses.

c) the draft proposes that only biometric identification systems should undergo a third-party conformity assessment whereas all the other “high risk” systems will be subject solely to self-assessment, which may possibly lead to serious safety concerns.

d) the lack of consumer protection provisions: the AIA does not establish rights for any affected AI user, to make any judicial claim or even lodge a complaint to the competent AI supervisory authority.

Overall, despite the various ambiguities that need to be addressed, the proposed AIA is undoubtedly a first step towards an integrated and coherent pan-European framework for the use and promotion of AI technology.

The editorial team