Legal provisions
Find out which regulatory provisions you need to observe when integrating AI technology.
AI in safety-related systems
Read how AI safety requirements can be determined using system analysis, risk management and statistical methods.
Accuracy level and confidence
Learn why confidence (trust) in the accuracy of AI is important.
What do accuracy and reliability mean in the context of AI?
What does the EU regulation on artificial intelligence (EU AI Act) require?
How can potential errors and risks be identified when integrating AI technologies?
How are target values for the required accuracy determined?
What options are there for determining target values for the confidence (trustworthiness) of the accuracy?
AI technologies are developing at a rapid pace and are increasingly being integrated into safety-related systems. In such cases, the EU regulation on artificial intelligence (EU AI Act) requires AI systems to be designed to achieve an “appropriate level of accuracy”.
This white paper outlines an approach to ensuring that this appropriate (or necessary) level of accuracy can be achieved. It first describes how to identify potential risks when integrating AI technologies into safety-related systems. We then show how the accuracy required to minimize these risks can be determined and the accuracy that can actually be achieved.
However, simply determining the achievable average accuracy is not enough for safety-related systems. It is also necessary to determine how high the confidence is that this accuracy will actually be achieved. The white paper provides you with a method for statistically determining the confidence - and thus for evaluating the suitability of an AI technology for integration into a safety-related system.
