Advancing Global Governance for Frontier AI: A Proposal for an AISI-Led Working Group under the AI Safety Summit Series

The rapid development of frontier AI models, capable of transformative societal impacts, has been acknowledged as an urgent governance challenge since the first AI Safety Summit at Bletchley Park in 2023 [1]. The successor summit in Seoul in 2024 marked significant progress, with sixteen leading companies committing to publish safety frameworks by the upcoming AI Action Summit [2]. Despite this progress, existing efforts, such as the EU AI Act [3] and voluntary industry commitments, remain either regional in scope or insufficiently coordinated, lacking the international standards necessary to ensure the universal safety of frontier AI systems.

This policy recommendation addresses these gaps by proposing that the AI Safety Summit series host a working group led by the AI Safety Institutes (AISIs). AISIs provide the technical expertise and resources essential for this endeavor, ensuring that the working group can develop international standard responsible scaling policies for frontier AI models [4]. The group would establish risk thresholds, deployment protocols, and monitoring mechanisms, enabling iterative updates based on advancements in AI safety research and stakeholder feedback.

The Summit series, with its recurring cadence and global participation, is uniquely positioned to foster a truly international governance effort. By inviting all countries to participate, this initiative would ensure equitable representation and broad adoption of harmonized global standards. These efforts would mitigate risks such as societal disruption, security vulnerabilities, and misuse, while supporting responsible innovation. Implementing this proposal at the 2025 AI Action Summit in Paris would establish a pivotal precedent for globally coordinated AI governance.