NEWSNews

  • 2025.02.11 NEW

    SAFE Project Outcomes - Results from the AI Action Summit Side Event in Paris -

    On February 10-11, 2025, the AI Action Summit was held at the Grand Palais in Paris. The summit brought together world leaders, international organizations, CEOs, academics, and civil society representatives to explore the scientific foundations, solutions, and standards for sustainable and beneficial AI.

    On February 11, as an official side event of the summit, the French GPAI Expert Support Center, Inria, hosted a meeting titled "Science and Technology Initiatives to Foster Cooperation... Time for Action" at Paris Santé Campus. The event covered key topics such as challenges in the new GPAI 2.0 integrated with the OECD, the outcomes of the SAFE project, the launch of a new project on Agentic AI, and ways to engage student communities in international AI discussions.

    1. SAFE Project Outcomes

    The event opened with a panel discussion featuring Ms. Karine Perset (Head of GPAI at OECD), Aurelie Simard (Executive Director of Inria, France), Ms. Sophie Fallaha (Executive Director of CEIMIA, Canada), and Dr. Harayama Yuko (Secretary General of the Tokyo ESC). They discussed the challenges in the new GPAI 2.0 integrated with the OECD, emphasizing the importance of collaboration with various stakeholders and ensuring representation.

    Following this, the Tokyo ESC presented the outcomes of the 2024 SAFE project. Co-Lead Cyrus Hodes and Associate Professor YOSHINAGA Kyoko from Keio University gave presentations. Co-Lead Hodes shared the results of Track C, which mapped the safety of AGI, while Professor Yoshinaga presented the results of Tracks A and B, focusing on technical safety and data governance. The database and website created in Track C, which organizes various technologies and solutions for AI risks from over 500 organizations worldwide, will also be published on the OECD website.

    2. Agentic AI Session

    The Tokyo ESC is set to lead a project on Agentic AI in 2025 and held a session to deepen understanding on this.

    In this session, Professor Vincent Corruble from Sorbonne University first provided an overview of Agentic AI, its historical background, and potential risks. The subsequent discussions emphasized the importance of safety, regulation, ethical considerations, and the need for control mechanisms. Experts also explored the possibility of Agentic AI acting autonomously, setting goals, and learning. The session also highlighted the social impact, the necessity for responsible development and governance, and the importance of balancing autonomy between humans and AI, drawing strong interest from the participating experts.

    Through these discussions and initiatives, under the GPAI 2.0 framework, the Tokyo ESC will deepen collaboration with diverse stakeholders and promote concrete actions to foster international cooperation while ensuring AI safety and ethics.

    Panel on GPAI 2.0 Left: Secretary General Harayama

    Panel on GPAI 2.0
    Left: Secretary General Harayama


    Explanation of Track C by Co-Lead Cyrus Hodes

    Explanation of Track C by Co-Lead Cyrus Hodes