AI Safety Papers

The Role of AI Safety Institutes in Contributing to International Standards for Frontier AI Safety

Authors

Kristina Fort

Abstract

International standards are crucial for ensuring that frontier AI systems are developed and deployed safely around the world. Since the AI Safety Institutes (AISIs) possess in-house technical expertise, mandate for international engagement, and convening power in the national AI ecosystem while being a government institution, we argue that they are particularly well-positioned to contribute to the international standard-setting processes for AI safety. In this paper, we propose and evaluate three models for AISI involvement: 1. Seoul Declaration Signatories, 2. US (and other Seoul Declaration Signatories) and China, and 3. Globally Inclusive. Leveraging their diverse strengths, these models are not mutually exclusive. Rather, they offer a multi-track system solution in which the central role of AISIs guarantees coherence among the different tracks and consistency in their AI safety focus.

Publication Details

  • Published: Unknown
  • Added to AI Safety Papers: 2024-11-13
  • Original Paper: Link

Summary

This paper examines the potential role of AI Safety Institutes (AISIs) in developing international standards for frontier AI safety. The authors propose three distinct but complementary models for AISI involvement in standard-setting processes, analyzing their respective strengths and limitations.

Key Points:

  1. Context and Motivation
  • International standards are crucial for ensuring safe development and deployment of frontier AI systems
  • AISIs are well-positioned to contribute due to their technical expertise, international engagement mandate, and convening power
  • Current standardization processes face challenges in keeping pace with rapid AI development
  1. Three Proposed Models

The paper presents three models for AISI involvement in standard-setting:

a) Model 1: Seoul Declaration Signatories

  • Includes original Seoul Declaration countries (Australia, Canada, EU, France, Germany, etc.)
  • High responsiveness and expertise, but lower legitimacy
  • Best suited for standards reflecting shared values among like-minded actors
  • Examples: safety thresholds, process standards

b) Model 2: US (+ Seoul Declaration Signatories) and China

  • Focuses on US-China cooperation with possible inclusion of Seoul Declaration signatories
  • Medium responsiveness and legitimacy, high expertise
  • Suited for standards fostering mutual trust and safety baselines
  • Examples: incident response plans

c) Model 3: Globally Inclusive

  • Reflects current ISO/IEC AI subcommittee membership
  • Lower responsiveness and expertise, but high legitimacy
  • Best for standards requiring broad international participation
  • Examples: interface standards, interpretability methods
  1. Analysis Framework The authors evaluate each model using three criteria:
  • Responsiveness: ability to keep pace with AI development
  • Legitimacy: breadth of stakeholder involvement and international acceptance
  • Expertise: technical capability to develop meaningful standards
  1. Key Findings As shown in Table 1, each model has distinct advantages and trade-offs. The authors suggest implementing these models as a multi-track system, with AISIs ensuring coherence across tracks and maintaining focus on AI safety.

The paper’s most important visualization is Table 1, which clearly summarizes the characteristics and suitable applications of each model. This table would make an excellent thumbnail as it captures the paper’s core contribution in a single, comprehensive view.

  1. Implications The research suggests that rather than choosing a single approach, a multi-track system leveraging all three models could provide the most comprehensive framework for international AI safety standards. The key is maintaining coherence through AISI involvement across all tracks.

This work makes a significant contribution to understanding how international AI safety standards might be developed effectively, balancing the need for technical expertise, international legitimacy, and rapid response to emerging challenges.