TÜV Austria presents new white paper “Safe and Certifiable Ai Systems”

In collaboration with the Johannes Kepler University Linz (JKU), the software Competence Center Hagenberg (SCCH) and the Joint Venture Trustifai, a continuous audit catalog was developed, which translates technical, legal and ethical requirements into verifiable criteria.

The Trusted Ai Framework of the TÜV Austria is based on three central pillars: safe software development, functional requirements as well as ethics and data protection. It offers a structured methodology for evaluating the functional trustworthiness of machine learning systems-including risk-based minimum requirements and statistical tests with independent data samples.

“With the Trusted Ai Framework, we create the basis for a transparent and reproducible certification of AI systems for the first time-a crucial step for their safe use in critical areas such as medicine, industry or mobility,” explains Andreas Gruber, Managing Director of TÜV Austria Joint Ventures Trustifai.

Target group: Responsible for the implementation of the requirements of the European KI regulation

TÜV Austria addresses the white paper specifically to those responsible in the areas of system development, corporate governance, quality and risk management. “These people are responsible for the implementation of the regulatory requirements and therefore play a key role in the successful implementation of AI,” continued Gruber.

The white paper not only documents the methodology, but also practical experience from the use of the test catalog. It shows typical sources of error such as data leaks, inadequate domain definitions or inadequate tests of AI applications and offers concrete recommendations for action.

“Our experience from many customer projects shows that companies need concrete requirements and implementation aids for the development and operation of AI systems. In our research cooperation with JKU and SCCH, with our white paper we offer a comprehensive insight into which processes, procedures as well as technical and organizational measures should be implemented as state-of-the-art,” says TÜV Austria CEO Stefan Haas.

With the Trusted Ai Framework, TÜV Austria delivers a practical timetable for providers, users and regulatory authorities-towards legally compliant, functionally trustworthy and certifiable AI systems in harmony with European standards.

“The regulation of AI must not only be based on principles-it must be verifiable in practice. Our framework connects technical best practice with the requirements of the EU KI regulation and makes trust measurable,” emphasizes Andreas Gruber.