Fraunhofer Institute for Cognitive Systems IKSIn the mechanical engineering standards landscape, application-specific standards are generally superior to generic or general standards. However, in the absence of an established application-specific solution, process, or methodology, these specific standards reference higher standards. In the area of artificial intelligence (AI) and machine learning (ML), this situation becomes even more complex. For example, some standards include aspects of AI and ML, while others do not.
© iStock.com/baranozdemir
Standards for AI, Safety and Automation
Challenge: Implementation and Compliance with Industry Standards
One problem is that not all relevant security standards take existing AI/ML solutions (or applications) into account. Verification of conformance is therefore complex and requires a customized approach. In addition, application-specific norms and standards related to AI/ML often refer to generic norms that specify the objectives to be achieved, but do not provide specific ways or methods to achieve them (using AI). This leads to another challenge in implementing and complying with security-related standards.
Whitepaper: The European Union Artificial Intelligence Act
In this whitepaper, we demonstrate how our methodology can be applied across three industry-specific usecases:
- Automotive: Automated Parking System
- Industrial Automation: Quality Inspection Cobot
- Healthcare: Brain-Computer Interface
The flexibility of the methodology allows for extensions and modifications depending on the use-case at hand. For each use-case, we present the approach for verifying a different sub-quality attribute for a particular stakeholder in the AI value chain.
Download Whitepaper: The European Union Artificial Intelligence Act