GPT-4o's audio output is restricted from generating erotic or violent speech content; the moderation model runs over a text transcription of the audio input and blocks generation if erotic or violent content is detected in the request.
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 input-side moderation mechanism for audio content that blocks erotic and violent speech generation based on transcription analysis, creating a disclosed content restriction applicable to GPT-4o's voice modality.
Interpretive note: The operational effectiveness of the transcription-based moderation mechanism is not fully characterized in the document for audio modality inputs, and the interaction with the disclosed audio robustness limitations creates uncertainty about consistent enforcement.
Under this provision, audio requests containing erotic or violent content will be blocked before generation, with the moderation classifier applied to the transcription of the audio input rather than the audio signal directly, meaning that the restriction is operationally dependent on accurate transcription of the audio input.
Cross-platform context
See how other platforms handle Erotic and Violent Speech Output Restriction and similar clauses.
Compare across platforms →"GPT-4o may be prompted to output erotic or violent speech content, which may be more evocative or harmful than the same context in text. Because of this, we decided to restrict the generation of erotic and violent speech. We run our existing moderation model over a text transcription of the audio input to detect if it contains a request for violent or erotic content, and will block a generation if so.Excerpt from OpenAI's GPT-4o System Card (PDF)
(1) REGULATORY LANDSCAPE: This provision engages communications decency and harmful content regulations, including the FTC Act's prohibition on unfair or deceptive practices in consumer-facing AI applications.
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 input-side moderation mechanism for audio content that blocks erotic and violent speech generation based on transcription analysis, creating a disclosed content restriction applicable to GPT-4o's voice modality.
Under this provision, audio requests containing erotic or violent content will be blocked before generation, with the moderation classifier applied to the transcription of the audio input rather than the audio signal directly, meaning that the restriction is operationally dependent on accurate transcription of the audio input.
No. ConductAtlas is an independent monitoring service. We are not affiliated with, endorsed by, or sponsored by OpenAI.