Key Outcomes of the GPAI Plenary - Tokyo Center to Lead Agentic AI Project

On April 8, 2025, the OECD hosted the Global Partnership on AI (GPAI) Plenary in an online format. The meeting brought together GPAI member countries and expert communities, including Expert Support Centers and individual experts, to discuss GPAI’s activity plans for 2025. From the GPAI Tokyo Expert Support Center (ESC), Dr. HARAYAMA Yuko, Secretary General of the Tokyo ESC and others participated in the discussions.

Following the Plenary, it was agreed that the Tokyo ESC would take the lead on a new project focused on Agentic AI systems, in collaboration with the OECD, starting in 2025.

Agentic AI systems, which can make autonomous decisions and take actions to achieve specific goals, are expected to boost productivity. However, they also raise concerns about potentially unpredictable real-world impacts. Given that Agentic AI is becoming a central theme in global discussions on the impact of AI, and in alignment with initiatives such as the Hiroshima AI Process’s Code of Conduct, the Tokyo Center’s project aims to develop a framework for analyzing the implications - both benefits and risks - of these systems, and classifying them based on their characteristics.

The OECD also announced that the Tokyo Center will host the Tokyo Innovation Workshop on May 28-29. This workshop will serve as a forum for discussing future proposals for GPAI projects and related initiatives for 2026.

Secretary General Harayama expressed her hope that the Tokyo Innovation Workshop would attract a diverse group of participants, including government officials and experts from both GPAI and non-GPAI countries. She also emphasized the importance of active and meaningful discussions in shaping stronger and more impactful projects for 2026.

The Plenary also included updates on GPAI’s major events in 2025, such as the GPAI Summit and upcoming Plenary meetings. Representing the Japanese government, the Ministry of Internal Affairs and Communications (MIC) encouraged participation in a new “Reporting Framework” under the Hiroshima AI Process. This framework invites AI developers and companies to voluntarily assess and disclose their adherence to the International Code of Conduct.