OpenAI's Approach to Frontier Risk (An Update for the UK AI Safety Summit)

• Authority: Private-sector companies
• Collections: Corporate policies and commitments
• Most recent activity: Enacted
• Most recent activity date: 2023-10-26

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
• Disinformation, surveillance, and influence at scale
• Cyberattacks, weapon development or use
• Fraud, scams, and targeted manipulation
• Governance failure
• AI pursuing its own goals
• AI possessing dangerous capabilities
• Lack of capability or robustness

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

• Exposure to toxic content
• Compromise of privacy
• Competitive dynamics
• Lack of transparency or interpretability

All other risk subdomains had no coverage in this document.

Risk Coverage Summary

The document has good coverage of approximately 10-12 subdomains, with strong focus on malicious actors (4.1, 4.2, 4.3), AI system security (2.2), dangerous capabilities (7.2), goal misalignment (7.1), lack of robustness (7.3), competitive dynamics (6.4), and governance failure (6.5). Coverage is concentrated in security, misuse prevention, and AI safety domains.

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 is an internal corporate policy document from OpenAI, an AI development company. The sectors governed are those in which OpenAI operates: Information (AI/software development) and Scientific Research and Development Services (AI research). The document does not regulate external sectors but describes OpenAI's own governance of its frontier AI development activities.

Legislative Status

Other

Legislative Status Summary

This is an internal corporate policy document describing OpenAI's voluntary commitments and preparedness framework for frontier AI risk management. It lacks binding legal obligations, enforcement mechanisms, or penalties.

Actors

Proposer

• Entity Types: AI_Developer
• Entities: OpenAI
• Summary: OpenAI is the author and proposer of this preparedness framework, describing their own voluntary commitments and internal governance structures for frontier AI development.

Target

• Entity Types: AI_Developer
• Entities: OpenAI; Microsoft
• Summary: The framework applies to OpenAI's own frontier model development and deployment processes. Microsoft is mentioned as a technology partner subject to the joint Deployment Safety Board.

Enforcer

• Entity Types: AI_Developer
• Entities: OpenAI Preparedness team; Superalignment team; Safety Systems team; Deployment Safety Board (OpenAI-Microsoft joint)
• Summary: Internal OpenAI teams and the joint Deployment Safety Board with Microsoft are responsible for implementing and enforcing the framework's provisions.

Monitor

• Entity Types: AI_Developer, AI_Governance_Actor
• Entities: OpenAI Preparedness team; Alignment Research Center (ARC); external red-teamers; Frontier Model Forum
• Summary: OpenAI's internal Preparedness team conducts ongoing monitoring, supplemented by external organizations like ARC for evaluations and the Frontier Model Forum for information sharing.

AI Lifecycle Stages

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

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

AI Lifecycle Summary

The document covers multiple AI lifecycle stages with primary focus on Build and Use Model, Verify and Validate, Deploy, and Operate and Monitor. It emphasizes evaluation, testing, deployment decisions, and post-deployment monitoring of frontier AI models.

Technical Scope

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

• AI Models
• Frontier AI
• AI Systems
• Generative AI

Technical Scope Summary

The document explicitly focuses on frontier AI models and systems, with detailed discussion of model capabilities, evaluations, and risks. It does not use the terms 'general purpose AI' or 'task-specific AI' but focuses on increasingly capable frontier models. No specific compute thresholds are mentioned. The document addresses model weights and their protection but does not explicitly discuss open-weight or open-source models.