PILLAR – 2
quality and testing
standards
In order to create planning security in the development of AI applications and to strengthen confidence in the use of AI applications, we need transparent and uniform AI quality and testing standards in line with European values and taking into account the requirements of the AI Act.
Background: AI Act and conformity assessment
Companies are obliged to assess their AI systems and the associated risk at the latest when the AI Act comes into force. Through a suitable conformity assessment, companies must prove that the requirements for trustworthy AI are met.
Voluntary AI quality standard
In cooperation with leading testing and standardisation institutions, MISSION KI is developing and testing a voluntary AI quality standard that represents a minimum standard of AI quality for specific applications. The AI quality standard consists of criteria (e.g. for trustworthy AI) as well as test procedures and test methods. Dimensions of the quality standard are, for example, transparency, reliability, fairness, data protection and management, and cybersecurity.
Norms and standards should help to ensure that AI systems are trustworthy, safe, inclusive and sustainable. They define the safety requirements that AI systems must fulfil before they come onto the market – for example in terms of transparency, accuracy, explainability or quality.
The voluntary AI quality and testing standard is being developed in four work packages:
PACKAGE – 1
Selection of suitable use cases
The first step is to identify real AI systems as use cases that offer broad sectoral and horizontal coverage.
PACKAGE – 2
Conception and development of the AI test procedure
At the same time, a comprehensive test procedure including test criteria, various test depths, technical test methods and an evaluation scheme is being developed.
PACKAGE – 3
Testing and further development of the AI test procedure
In order to establish the test procedure in the long term and to gather experience at an early stage, which will enable it to be refined, it is being trialled on the basis of use cases.
PACKAGE – 4
Establishment of a voluntary quality standard
From 2025, the consolidation of the test procedure and the voluntary quality promise for AI applications will be driven forward by anchoring it in committees and standards.
Focus and use cases
The MISSION KI quality standard focuses on AI applications that are below the high-risk cases (classification according to the European AI Regulation), but its content will also be compatible with the requirements of the AI Regulation for high-risk AI applications. In order to develop a marketable AI quality and testing standard, a use case-based procedure will be used for testing. For this purpose, use cases are currently being identified that offer broad sectoral and horizontal coverage. In order to account for the further development of AI applications and the associated, potentially increasing risk, high-risk AI applications are also included
Objective and outlook
The conformity assessment bodies can implement the test standards in their independent certifications. This makes an important contribution to ensuring that users can trust AI systems and use them with confidence. This strengthens the marketability of AI applications and thus the competitiveness of German companies.
Our partners
For this project, MISSION KI is pooling the expertise of leading testing and standardisation institutes:
INNOVATION & QUALITY CENTRES
MISSION KI is building innovation and quality centres that create a testing and experience environment for trustworthy AI. The first centres are located in Berlin and Kaiserslautern. The AI centre in Berlin is aimed at anyone who wants to find out more about AI and try out applications. It will be a place to meet and engage in dialogue about how AI should be used and within what limits. In Kaiserslautern, MISSION KI wants to create a contact point for founders, start-ups, SMEs and large companies that want to transfer AI from research to application. The AI Centre Kaiserslautern gives them access to cutting-edge research and AI testing environments – thanks to a cooperation with the German Research Centre for Artificial Intelligence (DFKI), which is based there.
News
News
News
News
CONTACT
NEWSLETTER
Sign up for our newsletter and be the first to hear about exciting events and progress at MISSION KI.