Development of a Toolkit on AI Safety

The GPAI Tokyo Expert Support Center has developed a toolkit on the safety of AI, in collaboration with the Mohammed Bin Rashid School of Government (MBRSG) in the United Arab Emirates (UAE), as part of the “SAFE Project” on generative AI safety and assurance, implemented under GPAI’s Work Plan 2024. This toolkit is characterized by its clear and detailed organization of risks and countermeasures related to Artificial General Intelligence (AGI), which have been examined by more than 500 organizations worldwide.

Moving forward, after necessary coordination with the OECD, the toolkit is scheduled to be published on the OECD’s website from July onwards.

1. Background

In the “G7 Leaders’ Statement on the Hiroshima AI Process” agreed upon in October 2023, as well as in other high-level government documents, G7 governments committed to advancing project-based cooperation with organizations such as GPAI to ensure the safety of generative AI.

Furthermore, in the Work Plan 2024 decided at the GPAI Summit held in India in December 2023, the launch of the SAFE Project was explicitly included as a new initiative informed by the Hiroshima AI Process.

Based on this Work Plan, the GPAI Tokyo Expert Support Centre has promoted the project since July 2024, working with MBRSG to develop the toolkit on the safety of general-purpose AI. In February 2025, the outcomes of the project were presented by Co-Chair Cyrus Hodes at a side event of the AI Action Summit held in Paris.

2. Achievements

This toolkit features a clear and detailed organization of risks and countermeasures related to general-purpose AI, as examined by more than 500 organizations worldwide.

Until now, there had been no single toolkit that comprehensively organized AI safety measures, creating obstacles for AI stakeholders such as policymakers, entrepreneurs, investors, researchers, and experts who aim to prioritize policy, regulation, and funding allocation. To fill this gap, the toolkit offers a structured inventory of AI risks, a catalog of existing technical and governance-based solutions to mitigate those risks and others.

As AI systems continue to grow in capability and scope of deployment, this mapping provides a critical foundation to ensure these technologies are developed and operated safely and responsibly.

3. Future Outlook

In relation to the development of this toolkit, a memorandum of understanding (MoU) is scheduled to be signed with MBRSG this April in the earliest, following the clarification of rights and cost-sharing related to the development.

Furthermore, after necessary coordination with the OECD regarding management and operations, the toolkit is scheduled to be published on the OECD’s website from July onwards.

The GPAI Tokyo Expert Support Centre will continue to promote international partnerships in the AI field, including through initiatives such as the “International Guiding Principles for Advanced AI Systems Developers” and the “International Code of Conduct for Advanced AI Systems Developers” under the Hiroshima AI Process.

(*) For a pre-site version, see MBRSG’s site: https://aisafetypreparedness.com/