How will the AI Act revolutionise the use of artificial intelligence?

On December 8, 2023, representatives of the European Commission, the European Parliament and member states reached a historic agreement on the final (though still subject to possible changes during the legislative process) shape of the EU's Artificial Intelligence Act (AI Act) regulation, the world's first comprehensive regulation setting rules for the use of artificial intelligence (AI) by both private and public entities. According to the draft's explanatory memorandum, the main goal of the document is to ensure the safe development of artificial intelligence and minimize the risk of abuse in the development and use of AI-based technology. 

Who is affected by EU AI Act? 

The regulation uses the term "artificial intelligence systems," which are software designed to operate with varying levels of autonomy that may exhibit adaptiveness after deployment and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence both physical or virtual environments. 

The EU AI Act will apply to so-called operators, by which is meant both providers, as well as deployers (users), importers and distributors of artificial intelligence systems. The above applies to any natural or legal persons, public authorities, agencies and other entities that develop artificial intelligence system or that has an artificial intelligence system developed their development (providers), use artificial intelligence systems (deployers), places on the market artificial intelligence systems (importers) or make artificial intelligence systems available on the EU market without being a provider or importer (distributors). 

It is worth noting that some exemptions are provided for the application of the EU AI Act and these will apply to artificial intelligence systems made available under free source and open source licenses, unless they are released as high-risk AI systems, which will be discussed later in this article. 

Prohibited AI practices in the European Union 

The new regulation, of course, involves the identification of prohibited artificial intelligence practices. Under the EU AI Act, it will be prohibited to place on the market, put into service or use an artificial intelligence system that employs subliminal techniques, exploits the vulnerabilities of a specific group of individuals, is designed to score or classify the credibility of individuals by public authorities (so-called "scoring" of citizens being a well-known and common practice in the People's Republic of China), and biometric identification of individuals "in real time" for the purpose of applying the law by government forces (with certain exceptions, such as for searching for victims of crime or countering terrorist acts). 

High-risk AI systems 

An important part of the EU AI Act will be dedicated to so-called high-risk artificial intelligence systems, related to sensitive areas specifically indicated in the regulation, including identification and biometric categorization of individuals, management of critical infrastructure, vocational education and training, hiring and management of employees, law enforcement, migration management, asylum and border control, administration of justice, and - of particular importance - access to essential private services and public services and benefits (e.g., health care, access to utilities, life and health insurance). 

If a particular artificial intelligence system meets high-risk criteria, then a number of additional requirements will need to be met, including testing, security, registration, compliance assessment and reporting obligations. 

Information obligations related to the use of artificial intelligence 

Providers of artificial intelligence systems designed to interact with individuals will be required to design and develop systems in such a way that individuals are informed that they are interacting with the artificial intelligence system. 

The principle of transparency with regard to artificial intelligence systems will also apply to their deployers (users), which in the case of an AI system that generates or manipulates image, audio or video content constituting a deep fake, shall disclose that the content has been artificially generated or manipulated. 

What is more, deployers (users) of an AI system that generates or manipulates text which is published with the purpose of informing the public on matters of public interest shall disclose that the text has been artificially generated or manipulated. All of the above mentioned obligations shall not apply where the use is authorised by law to detect, prevent, investigate and prosecute criminal offence. 

When will EU AI Act become effective? 

The EU AI Act is expected to pass through the legislative process in the coming months. The regulation shall apply from 24 months following the entering into force. During this two-year period, entities using artificial intelligence systems will have to prepare to apply the new requirements. What is more, European Agency for Artificial Intelligence will be established to support member states and their authorities.