This presentation gives an overview and further reading for the EU AI Act – drafted in February 2021, has been extensively discussed in the EU Council and Parliament is due for final agreement in late 2023, and will be enforced by the end of 2025.

The AI Act is based on broard horizontal classifications of risks, focusing on steps to mitigate the highest risks and forbid some use cases. The goal is to protect human rights, public safety, and environment, from misuse, unreliable operation, or bias of an AI system.

Reading the AI ACT, the preliminaries explain the legal basis, scope, and concepts, then there are sections (Articles) with specific requirements on risk management, data quality, oversight, operational monitoring, documentation, etc.

Requirements are most strict for high-risk scenarios, which are broadly defined but subject to case-by-case review by national Notified Bodies before deployment.

AI producers and deployers are subject to significant fines for breaking the rules and thereby harming EU citizens, no matter where in the world they are (similar approach to GDPR) . Harmonised standards are under development to operationalize the rules, for product conformance.

Download the presentation at the link below;