Joint Statement on Risk Assessment of Advanced AI Systems (International Network of AI Safety Institutes)

• Authority: Other multinational
• Jurisdiction: Multinational
• Collections: Multinational
• Most recent activity: Enacted
• Most recent activity date: 2024-11-20

Risk subdomain coverage

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

• 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.

• Compromise of privacy
• AI system security vulnerabilities
• Disinformation, surveillance, and influence at scale
• Cyberattacks, weapon development or use
• Fraud, scams, and targeted manipulation
• Competitive dynamics
• AI pursuing its own goals
• Multi-agent risks

All other risk subdomains had no coverage in this document.

Risk Coverage Summary

The document has minimal to good coverage of approximately 8-10 subdomains, with strongest focus on AI system safety failures (7.1, 7.2, 7.3), governance failure (6.5), and competitive dynamics (6.4). The document emphasizes risk assessment methodologies rather than specific risk types, with implicit coverage of malicious actors (4.1, 4.2, 4.3) and multi-agent risks (7.6).

Sector coverage

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

• Scientific Research & Development Services
• Public Administration (excl. National Security)

All other sectors had no coverage in this document.

Sectors Summary

This document does not govern specific economic sectors. It is a cross-sectoral framework for AI risk assessment methodology applicable to advanced AI systems across all sectors. The document references the Sustainable Development Goals, suggesting broad applicability, but does not establish sector-specific governance measures.

Legislative Status

Soft Law

Legislative Status Summary

This is a non-binding joint statement that establishes voluntary principles and commitments for AI risk assessment. It uses predominantly voluntary language ('should') and relies on collaborative adherence rather than legal enforcement mechanisms.

Actors

Proposer

• Entity Types: AI_Governance_Actor
• Summary: The document is explicitly authored and proposed by the International Network of AI Safety Institutes, which is a multinational network of government AI safety institutes committed to establishing shared scientific basis for AI risk assessments.

Target

• Entity Types: AI_Developer, AI_Governance_Actor
• Summary: The document targets network members who will conduct risk assessments, as well as organizations developing and deploying advanced AI systems. The guidance applies to those assessing risks of frontier AI systems, dual-use foundation models, general-purpose AI models, and advanced generative AI systems.

Enforcer

• Summary: The document does not specify any enforcement body or mechanisms. It is a voluntary commitment statement with no binding enforcement provisions. Individual network members retain flexibility to implement according to their own frameworks.

Monitor

• Entity Types: AI_Governance_Actor
• Summary: The document suggests collaborative monitoring through the Network's joint activities and mentions independent third-party evaluators for reproducibility. However, no formal monitoring body is designated.

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 iterative risk assessment throughout development, deployment, and post-deployment monitoring. It explicitly addresses planning, development, deployment, and operational monitoring stages, emphasizing that risks should be assessed and mitigated across all stages of the AI lifecycle.

Technical Scope

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

• AI Models
• Foundation Models
• Frontier AI
• AI Systems
• General Purpose AI
• Generative AI

Technical Scope Summary

The document explicitly mentions AI systems and multiple types of advanced AI including frontier AI systems, dual-use foundation models, general-purpose AI models, and advanced generative AI systems. It does not mention specific compute thresholds or distinguish between open-weight and closed models. The focus is on advanced AI systems broadly defined.