The policy update is dated October 21, 2026; the prior version remains in effect until the new date.
The requirement covers a model router, AI gateway, or another service that provides direct model access to third parties. The policy distinguishes these services from an application that uses OpenAI models to power its own product features. Integrating OpenAI models into an application does not, by itself, trigger the intermediary requirement.
The published material also updates language on using OpenAI models or outputs to train substitute models, protections for minors, high-stakes uses, and approval for national-security and law-enforcement uses. It does not establish which intermediary platforms have been approved, the approval criteria or review timeline, enforcement mechanics, affected providers, or commercial terms for gateway operators.