TÜV NORD CEO: New regulation will enable trustworthy AI

With the EU AI Act, Europe has finalized the world's first comprehensive AI law. “Regulation under the EU AI Act will make trustworthy artificial intelligence (AI) available that is human-centered and will be guided by democratic, ethical and moral principles,” says Dirk Stenkamp, CEO of TÜV NORD AG in the podcast ENTDECKT. EXPLAINED. NARRATED. With its regulation, the EU aims to cover three key points: safety, security and privacy. The riskier a system is, the higher the security requirements should be.

“In future, high-risk applications are to be checked by neutral third-party auditors such as TÜV NORD. These include biometric systems, which involve fingerprint recognition, or systems that are used in critical infrastructure, lending, human resources (analysis of CVs) and medical technology (e.g. surgical robots),” says Dr. Stenkamp. According to the regulation, unacceptable AI applications such as emotion recognition should not be used at all.

Test methods and criteria are required in order to be able to implement independent third-party testing by an AI TÜV. As soon as the EU publishes the legal text of the AI Act, standardization committees will start translating it into verifiable rules for products.

“As a neutral testing service provider, we in the TÜV NORD GROUP will apply the EU AI Act regulations to specific product tests. This can be software for speech recognition, which uses so-called large language models, or AI-based algorithms for autonomous vehicles, which we test according to defined criteria and standards,” says Dr. Stenkamp. There are essentially four basic criteria, such as the transparency of the algorithm or the deceptiveness of the AI. It is checked whether and how an AI algorithm can be tricked. Dr. Stenkamp also mentions biasing:

“It must be prevented that an AI makes biased or discriminatory decisions. Applied to the example of colors, this means that an AI must not prefer the color red over other colors.” Another important element of AI testing concerns reliability: “AI must produce consistent results. Put simply: given the same input, an AI-based algorithm must always deliver the same output,” says Dr. Stenkamp.

The TÜV NORD GROUP is highly committed to the development of test criteria and methods. “We have built up a lot of expertise in-house through TÜVIT experts. We work together with leading universities and research institutions and established an AI Lab years ago via the TÜV association. This has now been implemented in the form of AI Lab GmbH as an independent company with the leading TÜV companies as shareholders. AI Lab GmbH primarily develops test methods and procedures based on the norms and standards of the EU AI Act,” says Dr. Stenkamp.

The current podcast episode of ENTDECKT. EXPLAINS. ERZÄHLT. can be found here: Podcast - TÜV NORD | #explore (tuev-nord.de) and wherever podcasts are available. 

Founded over 150 years ago, we stand for security and trust worldwide. As a knowledge company, we have our sights firmly set on the digital future. Whether engineers, IT security experts or specialists for the mobility of the future: in more than 100 countries, we ensure that our customers become even more successful in the networked world.