Frontier AI Safety Commitments, AI Seoul Summit 2024

• Authority: Private-sector companies
• Collections: Corporate policies and commitments
• Most recent activity: Proposed
• Most recent activity date: 2024-05-21

Risk subdomain coverage

Good coverage — Comprehensive governance measures or mitigations with clear procedures relating to the risk subdomain are explicitly described.

• AI system security vulnerabilities
• Cyberattacks, weapon development or use
• Competitive dynamics
• Governance failure
• AI possessing dangerous capabilities
• Lack of capability or robustness
• Lack of transparency or interpretability

Minimal coverage — Brief mention but no specific mitigations, recommendations or governance measures relating to the risk subdomain are described.

• Pollution of information ecosystem
• Disinformation, surveillance, and influence at scale
• Fraud, scams, and targeted manipulation
• Overreliance and unsafe use
• AI pursuing its own goals

All other risk subdomains had no coverage in this document.

Risk Coverage Summary

The document has good coverage of approximately 8-10 subdomains, with strong focus on AI system security (2.2), malicious actors and misuse (4.1, 4.2, 4.3), competitive dynamics (6.4), governance failure (6.5), and AI safety failures (7.1, 7.2, 7.3). Coverage is concentrated in security, misuse prevention, and AI safety domains, with minimal attention to discrimination, privacy, misinformation, or socioeconomic impacts.

Sector coverage

Good coverage — Comprehensive governance measures or mitigations with clear procedures relating to the sector are explicitly described.

• Information
• Scientific Research & Development Services

All other sectors had no coverage in this document.

Sectors Summary

This document governs AI development and deployment across the Information sector and Scientific Research and Development Services sector, as it applies to frontier AI model developers. The commitments are sector-agnostic in their application, focusing on the organizations developing frontier AI rather than specific use cases in particular industries.

Legislative Status

Soft Law

Legislative Status Summary

This document represents voluntary commitments by private sector companies with no binding legal obligations, enforcement mechanisms, or penalties. The language is explicitly voluntary and relies on organizational self-governance and reputational incentives.

Actors

Proposer

• Entity Types: AI_Governance_Actor
• Entities: UK government; Republic of Korea government
• Summary: The document explicitly states that the UK and Republic of Korea governments announced these commitments and facilitated the agreement among private sector companies.

Target

• Entity Types: AI_Developer
• Entities: Amazon; Anthropic; Cohere; Google; G42; IBM; Inflection AI; Meta; Microsoft; Mistral AI; Naver; OpenAI; Samsung Electronics; Technology Innovation Institute; xAI; Zhipu.ai
• Summary: The document explicitly lists frontier AI model developers who have agreed to these commitments. These are organizations that develop highly capable general-purpose AI models.

Enforcer

• Entity Types: AI_Governance_Actor
• Entities: home governments; appointed body
• Summary: The document references home governments and appointed bodies as recipients of detailed information and participants in evaluation processes, though enforcement is primarily through transparency and self-governance rather than formal penalties.

Monitor

• Entity Types: AI_Governance_Actor, Affected_Stakeholder
• Entities: home governments; independent third-party evaluators; civil society; academics; the public
• Summary: The document specifies multiple monitoring actors including governments, independent evaluators, and broader stakeholder groups who are involved in assessing risks and adherence to the framework.

AI Lifecycle Stages

Lifecycle stages covered by the documents, as defined in the OECD Framework for the Classification of AI Systems

• Plan and Design
• Build and Use Model
• Verify and Validate
• Deploy
• Operate and Monitor

AI Lifecycle Summary

The document comprehensively covers the entire AI lifecycle with particular emphasis on risk assessment before and during training, deployment decisions, and ongoing monitoring. It explicitly addresses planning, development, evaluation, deployment, and operational monitoring stages.

Technical Scope

Terms relating to technical scope that are mentioned or implicitly referenced in the document

• AI Models
• Frontier AI
• AI Systems
• General Purpose AI

Technical Scope Summary

The document explicitly focuses on frontier AI models and systems, defining frontier AI as highly capable general-purpose AI. It does not mention specific compute thresholds, open-weight models, or distinguish between generative and predictive AI. The scope is clearly defined around frontier AI capabilities.