GRASP/SAFE Mapping Launch Event - Strengthening International Cooperation on AI Safety -

On February 9, 2025, the Mohammed Bin Rashid School of Government (MBRSG) in the UAE hosted the International AI Safety Cooperation & GRASP/SAFE Mapping Launch Event as a side event (AI Safety Connect 2025) of the AI Action Summit held in Paris. The event aimed to promote global collaboration in AI safety research and development, with Dr. HARAYAMA Yuko, Secretary General of the GPAI Tokyo Expert Support Center, moderating a roundtable discussion on AI safety institutions.

The event featured roundtables and panel discussions on international cooperation in AI safety the current state of AI safety institutions, frontier labs, AGI (Artificial General Intelligence), and safety science. In addition, there were also sessions for startups and investors, focusing on the current efforts and challenges in AI safety.

In particular, Professor Stuart Russell from UC Berkeley emphasized the need to establish a clear "behavioral red line" and implement ex ante regulation to ensure AI systems remain safe and beneficial for humanity. He stated, "Until now, safety and innovation have been seen as a trade-off, but innovation and development cannot exist without safety," which received strong agreement from the participants.

Professor Yoshua Bengio from the University of Montreal highlighted that as AGI develops into autonomous agents, the goals that AGI sets for their own activities may not necessarily align with human well-being and values. He warned that, in the worst case, this could lead to "self-destructive outcomes," and emphasized the need for governments to establish incentives to ensure accountability among developers of frontier AI models.

Through an event like this, the Tokyo Center will leverage opportunities to strengthen international cooperation and establish concrete action plans for AI safety. The Tokyo Center is committed to contributing to these efforts.