High-risk AI system

An AI system that the EU AI Act allows, but only under its strictest rules, because of what it is used for. A system falls into this group in one of two ways. The first covers AI that is a safety feature of a product, or is the product itself. It applies when EU safety rules require an independent body to check that product. The second covers AI used in one of eight listed areas, including jobs and the workplace, education, critical infrastructure and access to essential services such as credit. What the system is used for decides, so the same technology is high-risk in one case and not in another. In the listed areas, a system that only does a small routine or preparatory step is exempt, but one that builds profiles of people from their personal data always counts. For a high-risk system, the maker must manage its risks and document it. The organisation using it must put people in charge of overseeing it and keep the records it produces. Systems in the listed areas must also be registered with the authorities. These duties start in December 2027 for the listed areas and in August 2028 for regulated products.