OpenAI's Preparedness Framework evaluation classified GPT-4o's text persuasion capability as borderline medium risk and voice persuasion as low risk, based on pre-registered thresholds and a study of over 3,800 participants comparing AI and human-authored persuasive content on political topics.
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 the documented basis for GPT-4o's medium overall risk classification under the Preparedness Framework and discloses the methodology and findings of a structured persuasion study, which is operationally relevant for AI governance assessments and regulatory inquiries into AI influence on political opinion.
The document discloses that GPT-4o's text modality marginally exceeded the medium risk threshold for persuasion based on pre-registered thresholds, with AI-generated text exceeding human-written content in three of twelve instances, while the voice modality was classified as low risk and not found to be more persuasive than a human baseline.
Cross-platform context
See how other platforms handle Persuasion Risk Classification and Evaluation Disclosure and similar clauses.
Compare across platforms →"Based on pre-registered thresholds, the voice modality was classified as low risk, while the text modality marginally crossed into medium risk. The AI interventions were not more persuasive than human-written content in aggregate, but they exceeded the human interventions in three instances out of twelve. We found that for both interactive multi-turn conversations and audio clips, the GPT-4o voice model was not more persuasive than a human.Excerpt from OpenAI's GPT-4o System Card (PDF)
(1) REGULATORY LANDSCAPE: Persuasion risk in AI systems is relevant to FTC Act unfair or deceptive practices authority, the EU AI Act's prohibited practice of using subliminal techniques to materially distort behavior, and emerging election …
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 the documented basis for GPT-4o's medium overall risk classification under the Preparedness Framework and discloses the methodology and findings of a structured persuasion study, which is operationally relevant for AI governance assessments and regulatory inquiries into AI influence on political opinion.
The document discloses that GPT-4o's text modality marginally exceeded the medium risk threshold for persuasion based on pre-registered thresholds, with AI-generated text exceeding human-written content in three of twelve instances, while the voice modality was classified as low risk and not found to be more persuasive than a human baseline.
No. ConductAtlas is an independent monitoring service. We are not affiliated with, endorsed by, or sponsored by OpenAI.