OpenAI will require pre-approval for model routers and AI gateways

What is changing?

OpenAI is updating its usage policies to align safeguards across services that provide access to its models. The company expects similar protections whether customers use OpenAI directly, a hosting partner, or a reseller, and calls for providers to investigate violations and coordinate enforcement across services.

Under the updated policies taking effect on November 20, 2026, services that provide direct access to OpenAI models—such as model routers or AI gateways—must obtain OpenAI review and approval before offering that access. The announcement also introduces SafetyIDs, which are intended to help customers distinguish individual users during investigations without disclosing their identities to OpenAI.

OpenAI says restrictions will be based on documented conduct and risk and can be revisited when circumstances change. Because the announcement does not detail the technical approval workflow, integration teams will still need to monitor the updated policy text and provider-specific requirements.