First and foremost, all AI systems enumerated in Annex III are considered high-risk regardless of any other factors.
AI systems that are not in Annex III can also be classified as high risk, provided they satisfy the following conditions:
=> The AI system is to be used as a safety component of a product or the AI system is itself a product classified under Annex I
=> The product or AI system for which it is a safety component is required to undergo third party assessment before it is placed on the market or put into service.
-
Annex III AI systems can be not high-risk
An AI system in Annex III is not considered high-risk if it does not pose any significant threat to the health, safety or fundamental rights of natural persons including not deploying any prohibited practices.
The exception is applicable if the AI system meets any of the following criteria:
=> Performs a narrow procedural task
=> Improves result of previously completed human activity
=> Detects decision-making patterns or deviations from past patterns without replacing or influencing previous human assessment unless reviewed by a human
=> Performs a preparatory task for assessment relevant to the use cases in Annex III.
Exception: No Annex III AI system that profiles natural persons can be exempted.
-
Documentation of Annex III non-high risk systems
Any provider that considers an AI-system under Annex III is not high-risk must document such assessment.
=> The assessment must be executed before putting the AI system into market or putting it into service.
=> The documentation of assessment must be provided to the competent national authorities on their request.
=> The provider must register themselves and the AI system in the EU database for high-risk AI systems.