Preparedness Framework
OpenAI's system for evaluating AI model risks and safety before deployment.
The Preparedness Framework is OpenAI's internal system for checking AI models. It helps them make sure that new AI models are safe and secure before they are released to the public.
This framework involves a set of guidelines and evaluations. OpenAI uses it to assess potential dangers and safety issues, especially with advanced AI capabilities. For example, they might use it to check if a powerful new model could be misused for harmful purposes.
The Preparedness Framework is OpenAI's structured process for evaluating and mitigating risks associated with advanced AI models. It encompasses a suite of internal guidelines, safety assessments, and tiered safety measures designed to identify potential harms and ensure responsible deployment. This framework is crucial for managing the cybersecurity implications and broader societal impacts of increasingly capable AI systems.
Its application involves rigorous testing against various threat models and potential misuse scenarios. The framework aims to balance rapid development with robust safety protocols, particularly for models exhibiting critical capabilities.
OpenAI's internal discussions and public statements highlight a growing focus on safety and risk assessment as AI capabilities advance. The term appears in contexts discussing the pacing of development, the evaluation of new models like GPT-5.5 and Astra, and concerns about the safety culture within the company. This suggests a broader industry trend towards formalizing preparedness for increasingly powerful AI systems.
- An OpenAI safety insider calls the culture 'broken'OpenAI · The Rundown AI ·
- Pacing model development in an era of cyber-critical capabilitiesOpenAI · OpenAI News ·
- Responding to the next frontier of critical cyber capabilitiesOpenAI · OpenAI News ·
- GPT-5.5 System CardOpenAI · OpenAI News ·
- GPT-5.3-Codex System CardOpenAI · OpenAI News ·
- ChatGPT agent System CardOpenAI · OpenAI News ·
- Deep research System CardOpenAI · OpenAI News ·
- OpenAI o3-mini System CardOpenAI · OpenAI News ·
- GPT-4o System CardOpenAI · OpenAI News ·
- OpenAI ChatGPT
- OpenAI uses a Preparedness Framework internally to evaluate AI models for safety and security risks before deployment, as seen in discussions around models like GPT-5.5 and Astra.
- Safety Guidelines
- These are general principles for safe AI development, while t
- Red Teaming
- This is a method of testing AI systems for vulnerabilities, often a
- Compliance Audits
- These are external checks for adherence to regulations, where
The term and concept appear to originate from OpenAI's internal processes for managing the development and deployment of advanced AI models, particularly as they develop more capable and potentially risky systems.
While the Preparedness Framework itself is internal, you can observe its outcomes by exploring OpenAI's safety documentation for models like GPT-4o. Look for sections detailing safety assessments, potential risks, and mitigation strategies to understand the types of concerns addressed by such frameworks.
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- PromptRole-play a student struggling with math for AI tutor simulationSimulate a student's math struggles and uncertainty for AI tutor evaluation. · 5 min
- TemplateReusable workflow templates for healthcare teamsTeams can create and share templates for common workflows like documentation or prior authorizations. · 15 min
- TemplateTemplate for building an AI tool tailored to your workA structured approach to defining and building a custom AI tool for specific business needs. · 15 min
- TemplateEvals CSV prompt list for agent skillsA CSV file template for listing prompts used in evaluating agent skills. · 20 min
- HeuristicDouble coding agent task completion with docs linksProviding coding agents with a documentation link significantly increases their task completion rate. · 2 min
Questions people ask
- What is the main goal of the Preparedness Framework?
- The main goal is to ensure AI models are safe and secure before they are released, by evaluating potential risks and implementing mitigation strategies.
- Who uses the Preparedness Framework?
- The Preparedness Framework is an internal system used by OpenAI to assess its own AI models.
- Does this framework guarantee AI will never be misused?
- No system can guarantee zero misuse, but the Preparedness Framework is designed to significantly reduce risks and prepare for potential issues.
First seen on Subvolts Aug 8, 2024. Updated Oct 12, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.