What changed
OpenAI announced on October 21 that its updated Usage Policies will take effect November 20, 2026. They require prior OpenAI review and approval for intermediary platforms or other services that let third parties directly access OpenAI models. The announcement names model routers and AI gateways as examples.
The policy also explicitly says this requirement does not apply solely because an application integrates OpenAI models to power its features or functionality. That distinction matters: an AI-powered application is not automatically a model-access intermediary.
A service arrangement to check
OpenAI describes related work to align safeguards across its own services, partner-hosted access, and resellers. It says access partners must maintain consistent safeguards and investigate potential violations.
This is an announced future policy change, not a new model or a verified interruption of any particular service. The announcement and policy do not establish that any named gateway or application needs approval.
These two linked sources do not describe a submission process or review timetable.
Our take
For experienced technicians supporting a small business, begin with the actual path a request takes. Write down who supplies model access, who operates the application, and what the customer can do through it. A dashboard label or the word gateway alone is a poor basis for deciding how a service fits the policy.
If a vendor provides direct model access, ask it to explain its position on the November requirement and what customers should expect. If you operate an application, compare its real functionality with the written exception instead of assuming every AI feature requires a separate approval process.
Keep that response with your service documentation and review dependencies before a renewal or migration. This briefing reports the published policy and suggests questions; we have not sought approval, checked a provider account, or tested enforcement.