RatioLogo
Back

The AI Summit Series at a Crossroads

The international effort to govern frontier AI, born from ad-hoc summits at Bletchley Park and Seoul, faces a critical juncture as it prepares for Paris 2025. The initial momentum is at risk from "focus drift," threatening to dilute the core mission of managing urgent AI safety risks.

A coalition of over 30 experts from leading institutions warns that the current improvisational approach cannot keep pace with a technology evolving monthly, leaving the public vulnerable without a permanent forum for accountability.

The Core Proposal: A Two-Track Future

To preserve focus while expanding participation, experts propose a radical restructuring.

Track 1: Governance & Safety

  • Purpose: Manage "frontier risks" from the most powerful AI systems.
  • Participation: Technically limited to leaders from major AI labs and relevant national bodies.
  • Goal: Prevent the watering down of critical safety standards as the summit grows.

Track 2: Public Interest & Inclusion

  • Purpose: Address broad societal benefits, economic impacts, and global inclusivity.
  • Participation: Open to the expanding roster of nations (projected to grow from 28 to over 90).
  • Goal: Ensure all voices are heard on AI's global future without hampering urgent safety work.

Operational Reforms for Sustained Momentum

Moving beyond temporary "host-only" agendas is essential for long-term effectiveness.

The "Troika" Host System

  • Establishes a coordinated succession between past, present, and future summit hosts.
  • Creates a continuous 18-month roadmap to maintain momentum on safety commitments.
  • Requires a 2-year lead time for host selection to prevent "knowledge loss" between cycles.

Supporting Infrastructure

  • A "Hybrid/Incremental" Secretariat provides institutional memory without creating a slow bureaucracy.
  • Annual high-level summits are supported by interim technical meetings to match AI's rapid development pace.

Critical Challenges & Friction Points

Even the most robust blueprint must navigate complex political realities.

  • Defining Leadership: Determining what qualifies as an "AI-leading" nation is politically fraught.
  • Global Legitimacy: The potential exclusion of China from technical networks like the AISI Network (9 countries + the EU) remains a major point of contention.
  • Political Volatility: A fractured geopolitical landscape threatens consensus on any permanent structure.

The Path Forward: A Race Against Time

The report frames the next steps as urgent "reverse institutionalization"—building governance around a runaway technology.

Short-Term Horizon (0–6 months): The immediate priority is to lock in the next three host nations to secure the Troika system.

The Ultimate Goal: To construct a governance cage strong enough for a technology that is quickly outgrowing its bars.

Reference: Velasco, L., Trager, R., et al. (2025). The Future of the AI Summit Series. Oxford Martin School AI Governance Initiative Policy Memo. Published January 2025.