The TÜV has called for mandatory, independent safety audits for language models like ChatGPT before they are allowed to enter the market. Franziska Weindauer, head of the TÜV AI Lab, told Focus that involving reviewers only during the development process is insufficient. In her view, recently developed chatbots should instead undergo a compulsory assessment prior to publication.
The German government is also addressing this issue. Since the end of August, it has been establishing the AI Safety Institute (AISI) in Berlin. Weindauer emphasized the necessity of developing specific German capabilities, noting that the success of the AISI depends on its pragmatic and rapid establishment, attracting the right personnel, and clearly defining its mandate. She cautioned that building the AISI will not be simple, citing past experiences with establishing other institutions, and stressed the need to clarify its currently unclear organizational structure as quickly as possible.
Despite safety concerns, Digital Minister Karsten Wildberger (CDU) appealed for the acceleration of AI development in Europe. According to his statements to the magazine, “If we build AI models ourselves, we also have control over the questions of values and safety.” He asserted that Germany must play a much more active role in the future, as this serves as the opportunity to secure prosperity and competitive strength for the coming decades.
