The DLR Institute for AI Safety and Security and the Japan AI Safety Institute (AISI) are working on specific joint projects to improve AI safety.
The focus is on linking the DLR Data Quality Toolkit with the AISI AI evaluation tools to demonstrably increase the trustworthiness of AI systems.
Following two visits by Japanese colleagues to Germany, our institute made a return visit to Tokyo on 17 October to agree on specific next steps.
This international cooperation is helping to establish globally harmonised safety standards and evaluation methods for AI systems.
Cooperation between the DLR Institute for AI Safety and Security and the Japan AI Safety Institute (AISI) is gaining momentum. After two visits by Japanese experts to Germany in spring and summer 2025, the time had come for a return visit. On 17 October, our colleague Steffen Turnbull travelled to Tokyo to discuss potential areas of cooperation with Hiromu Kitamura and other AISI experts. The discussions in Akasaka focused on a promising approach: integrating DLR's Data Quality Toolkit with the AI evaluation tools of the Japanese partner institute. This combination is important because data quality and AI evaluation are inextricably linked. Even highly developed AI systems can become unreliable if the quality of the underlying data is poor.
International standards for global challenges
The meeting in Tokyo once again emphasised the importance of international cooperation in the field of AI safety. As AI systems are developed and used worldwide, it is crucial that safety standards and evaluation methods are harmonised across countries. Both institutes share the goal of making scientifically sound contributions to global AI governance, as well as developing methods that measurably and verifiably increase the trustworthiness of AI systems.
The visit was more than just a diplomatic exchange — concrete next steps were agreed to give the cooperation practical form. Despite the geographical distance between Germany and Japan, the two countries are united by their shared goal of making AI systems safer and more trustworthy.
