OpenAI Preparedness Framework (Beta)

• Authority: Private-sector companies
• Collections: Corporate policies and commitments;Editors' Picks
• Most recent activity: Enacted
• Most recent activity date: 2023-12-18

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.

• Competitive dynamics

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), 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 coverage of discrimination, privacy, misinformation, and 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 is an internal corporate policy document that governs OpenAI's own operations. As an AI development company, OpenAI operates primarily in the Information sector (AI/ML development, data processing) and Scientific Research and Development Services sector (AI research). The document does not regulate external sectors but rather establishes internal governance for OpenAI's AI development activities.

Legislative Status

Other

Legislative Status Summary

This is an internal corporate policy document establishing OpenAI's voluntary framework for managing catastrophic AI risks. It contains internal governance structures, safety baselines, and procedural commitments but lacks external enforcement mechanisms or legal penalties.

Actors

Proposer

• Entity Types: AI_Developer
• Entities: OpenAI
• Summary: OpenAI is the author and proposer of this Preparedness Framework, as evidenced by the document being titled 'OpenAI Preparedness Framework' and references throughout to 'we' and 'our' referring to OpenAI's processes and commitments.

Target

• Entity Types: AI_Developer
• Entities: OpenAI; Preparedness team; Safety Systems team; Security team; Superalignment team; Policy Research team
• Summary: The framework applies to OpenAI's internal operations, teams, and model development processes. It establishes requirements for various internal teams and governs OpenAI's own AI development and deployment activities.

Enforcer

• Entity Types: AI_Developer
• Entities: Safety Advisory Group (SAG); OpenAI Leadership; OpenAI Board of Directors; SAG Chair
• Summary: The framework establishes internal enforcement through the Safety Advisory Group, OpenAI Leadership, and Board of Directors, with the Board having ultimate oversight and ability to reverse decisions.

Monitor

• Entity Types: AI_Developer, AI_Governance_Actor
• Entities: Preparedness team; Safety Advisory Group (SAG); Trustworthy AI team; qualified independent third-parties
• Summary: The Preparedness team is responsible for ongoing monitoring, evaluation, and reporting. Third-party auditors provide independent verification, and the SAG oversees the assessment of the risk landscape.

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 comprehensively covers all AI lifecycle stages with particular emphasis on Build and Use Model, Verify and Validate, Deploy, and Operate and Monitor. It addresses planning through risk category identification, model development through evaluation processes, validation through pre- and post-mitigation testing, deployment restrictions based on risk scores, and ongoing monitoring of deployed models.

Technical Scope

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

• AI Models
• Frontier AI
• AI Systems
• Compute Threshold
• Open Weight / Open Source

Technical Scope Summary

The document explicitly focuses on 'frontier models' and 'frontier AI models' throughout, referring to increasingly powerful AI models approaching AGI. It does not use terms like 'general purpose AI', 'task-specific AI', 'foundation models', or 'generative AI' explicitly. The document mentions compute thresholds (>2x effective compute increase) as triggers for evaluation. There is one reference to model weights in the context of open release decisions.