OpenAI's Preparedness Framework establishes that models scoring above medium risk post-mitigation cannot be deployed, and models scoring above high risk post-mitigation cannot continue development, with GPT-4o classified as medium overall due to a borderline medium persuasion score.
This analysis describes what OpenAI's agreement states, permits, or reserves. It does not constitute a legal determination about enforceability. Regulatory applicability and practical outcomes may vary by jurisdiction, enforcement context, and individual circumstances. Read our methodology
This provision establishes an internal governance gate that defines the conditions under which a frontier model may be released, creating a documented procedural commitment that downstream API operators and enterprise customers may reference when assessing OpenAI's deployment governance practices.
This provision establishes that GPT-4o's deployment was authorized only after post-mitigation evaluations across cybersecurity, biological threats, persuasion, and model autonomy produced scores at or below the medium threshold, with persuasion identified as the category that marginally crossed into medium risk.
Cross-platform context
See how other platforms handle Preparedness Framework Deployment Threshold and similar clauses.
Compare across platforms →"If a model passes a high risk threshold, we do not deploy the model until mitigations lower the score to medium. Only models with a post-mitigation score of medium or below can be deployed. Only models with a post-mitigation score of high or below can be developed further.Excerpt from OpenAI's GPT-4o System Card (PDF)
(1) REGULATORY LANDSCAPE: This provision engages with the EU AI Act's requirements for general-purpose AI models with systemic risk potential, which include adversarial testing and risk assessment obligations.
Enforcement risk, jurisdiction flags, contract triggers, and due diligence action items.
Get the research letter
Companies change their terms quietly. We read every version and catch what actually changed. One email a week on the changes that matter and what they mean.
This provision establishes an internal governance gate that defines the conditions under which a frontier model may be released, creating a documented procedural commitment that downstream API operators and enterprise customers may reference when assessing OpenAI's deployment governance practices.
This provision establishes that GPT-4o's deployment was authorized only after post-mitigation evaluations across cybersecurity, biological threats, persuasion, and model autonomy produced scores at or below the medium threshold, with persuasion identified as the category that marginally crossed into medium risk.
No. ConductAtlas is an independent monitoring service. We are not affiliated with, endorsed by, or sponsored by OpenAI.