OpenAI Usage Policies

This document is OpenAI's Usage Policy (Acceptable Use Policy), governing permissible and prohibited uses of OpenAI's models, tools, APIs, and products, with its authority grounded in OpenAI's Terms of Use. The policy asserts that all users and operators must comply with a defined set of prohibited use categories and that operators building on the API are responsible for ensuring their end users also comply, effectively creating a two-tier compliance obligation. The policy enumerates absolute prohibitions — including generation of child sexual abuse material, creation of cyberweapons, development of weapons of mass destruction, and content facilitating real-world violence — alongside conditional restrictions where context, safeguards, and operator permissions can modify what is permissible, a structure that places significant interpretive and enforcement discretion with OpenAI. The policy engages with multiple regulatory frameworks including the EU AI Act (which classifies certain AI uses as prohibited or high-risk), COPPA and child safety statutes, computer fraud and cybercrime laws across jurisdictions, export control regimes (ITAR, EAR), and platform liability frameworks such as Section 230 of the CDA; the policy's operator responsibility provisions may create compliance surface area under each of these depending on use case and jurisdiction. Compliance teams deploying OpenAI via API should note that the operator tier carries downstream liability exposure for end-user violations, that certain permitted-by-default behaviors can be unlocked only by operators who meet unspecified eligibility criteria, and that OpenAI reserves unilateral authority to update usage policies without specifying notice obligations or effective-date timelines.