Usage Policies

This is OpenAI's Usage Policies document, governing what users and developers may and may not do with OpenAI's products including ChatGPT, Sora, Codex, and the API. The policy prohibits a defined list of activities including weapons development, real-time biometric identification in public spaces, social scoring, criminal risk prediction from personal traits, content sexualizing minors, and automation of high-stakes decisions in sensitive domains such as housing, employment, medical care, and law enforcement without human review. The policy also states that OpenAI reports apparent child sexual abuse material to the National Center for Missing and Exploited Children and reserves the right to revoke access and apply penalties for policy violations.

This document is OpenAI's Usage Policies, effective October 29, 2025, governing permissible and prohibited uses of OpenAI products and services including ChatGPT, Sora, Codex, and the API platform. The policy states that users may not use OpenAI services for threats, harassment, weapons development, CBRNE-related activity, malicious cyber activity, real money gambling, or national security and intelligence purposes without OpenAI review and approval; the terms further authorize OpenAI to withhold access where it reasonably believes doing so is necessary to protect the service, users, or third parties. The policy includes categorical prohibitions on real-time remote biometric identification in public spaces, social scoring and profiling, emotional inference in workplace and educational settings, and criminal risk prediction based solely on personal traits or profiling, which align with frameworks such as the EU AI Act's prohibited AI practices. The document engages EU AI Act prohibited practices provisions, COPPA-adjacent child safety obligations (including mandatory NCMEC reporting of apparent CSAM), and general consumer protection frameworks enforced by the FTC; applicability of specific regulatory frameworks depends on the jurisdiction in which a given user or operator is located. The October 2025 update states that the policy now reflects a universal set of policies across all OpenAI products and services, consolidating prior service-specific guidance.

The agreement establishes a universal set of prohibited use categories across all OpenAI products, including restrictions on generating content that threatens, harasses, or defames individuals, and on automating decisions in sensitive domains such as housing, employment, medical care, and law enforcement without human review. Under these terms, violations may result in loss of access to OpenAI systems or other unspecified penalties, as the policy reserves OpenAI's right to withhold access where it reasonably believes doing so is necessary. You can appeal enforcement actions if you believe OpenAI has made an error applying the policy.