GPAI SAFE project report: Technical trustworthiness and data governance assurance of models

The GPAI Project Safety and Assurance of Generative AI (SAFE) was launched as a cross-Working Group initiative based on the GPAI Work Plan 2024, in response to growing global concerns about the risks and opportunities posed by advanced artificial intelligence systems. Framed by the objectives set forth by the G7 Hiroshima AI Process and aligned with the commitments made at the Bletchley and Seoul AI Summits, this report addresses both the technical and ethical dimensions of Generative AI, with a focus on interoperability, validation, data governance, and societal impact. The project consists of two components: “Technical Trustworthiness and Data Governance Assurance of Models” and “Mapping AGI Safety Solutions.”