Symposia

AI continues to evolve rapidly. AI agents are becoming more capable, and expectations are growing for physical AI, including robots operating in the real world. AI development and use are becoming central to policy in Japan and foundational to society and industry.

At the same time, safety challenges are becoming tangible. Reports describe unexpected behavior during advanced model evaluations and cases in which excessive dependence on conversational AI has been linked to tragic incidents involving minors. Looking ahead, questions include how humanoid robots can be used safely in care and household settings, how experts should work with AI that exceeds human capabilities in particular domains, and how to handle generative AI outputs based on data involving copyright and personality rights.

Against this background, this symposium presents AIST’s research and development efforts toward AI safety.