Preparedness Framework
webCredibility Rating
High quality. Established institution or organization with editorial oversight and accountability.
Rating inherited from publication venue: OpenAI
The page returned a 404 at time of analysis; metadata is based on the known public document. The Preparedness Framework (published late 2023) is a key reference for understanding how OpenAI operationalizes safety evaluations for frontier models.
Metadata
Summary
The OpenAI Preparedness Framework is a policy document outlining how OpenAI evaluates and mitigates catastrophic risks from frontier AI models, including CBRN threats, cyberattacks, and loss of human control. It establishes a risk classification system and safety thresholds that determine whether models can be deployed or developed further. The framework is a key governance artifact representing OpenAI's approach to pre-deployment safety assessment.
Key Points
- •Defines a tiered risk scoring system (low, medium, high, critical) across risk categories including CBRN, cyberoffense, and autonomy/control
- •Establishes that models scoring 'critical' on any category cannot be deployed, and 'high' requires safety mitigations before deployment
- •Creates a dedicated Preparedness team responsible for tracking, evaluating, and forecasting frontier model risks
- •Introduces a 'Safety Baseline' requiring ongoing monitoring and the ability to course-correct if risks emerge post-deployment
- •Represents a formalized institutional commitment to pre-deployment risk evaluation at a leading frontier AI lab
Cited by 1 page
| Page | Type | Quality |
|---|---|---|
| Agentic AI | Capability | 68.0 |
Cached Content Preview
OpenAI # 404 The door missed its cue Backstage, pulleys hum along Try the velvet rope by [gpt-5.4-thinking(opens in a new window)](https://chatgpt.com/?model=gpt-5.4-thinking&openaicom-did=4989dbe5-4f1e-44f6-ae40-aba6c5df2268&openaicom_referred=true)
474033f678dfe09a | Stable ID: NTliYzc0Ym