If adopted, The Artificial Intelligence Act proposal will regulate the AI use for service providers and users in respect to the EU, thus resulting in the “Brussels effect”, where even the entities located and established outside the EU and EEA (i.e., “the third countries”) will have to observe the regulation terms, as long as they provide services for the EU/EEA territory or to its citizens. The regulation promotes the trustworthy use of AI systems, limiting the permitted AI uses by introducing the prohibitions of certain AI systems, for example, the use of biometric systems in public spaces, use of the emotion recognition and predictive policing systems.
Risk and Responsibility
The first version of The Artificial Intelligence Act introduced in 2021 proposed a risk based approach system, dividing the AI systems in three risk categories – unacceptable risk (prohibited, for example – social scoring systems, systems, that explore and target vulnerabilities of certain social groups, etc.), high risk (the use limited and strictly regulated, for instance – the systems used in civil aviation security, medical industry, operation of critical infrastructure, etc.) and the limited risk category (systems, that interact with humans directly – chatbots, AI translation systems, etc.), the latter being permitted, however, complying with the transparency duty – to inform the users, that the AI system is at use.
The key regulations of The Artificial Intelligence Act include the obligation of the AI systems service providers to disclose the AI generated content training data summary, the monitoring of cybersecurity, ensuring that the fundamental rights of the system users are protected, as well as informing of the AI software users that the AI is used. Thus, some of the service provider obligations overlap with the Digital Services Regulation (Regulation (EU) No. 2022/2065 of the European Parliament and of the Council of 19 October, 2022), creating a uniform EU Digital single market regulation within the EU.
Similarly to the competition law breaches the AI regulation requirements are enforced by administrative fines system – up to 40 million EUR, or 7 % of the total worldwide annual turnover for The Artificial Intelligence Act breach.