OpenAI is adding a bouncer to the door. Under its refreshed Usage Policies, operators of model routers and AI gateways now need prior review and approval before offering third-party access to OpenAI’s models.
The new rules take effect on November 20, 2026. The policy revision itself is dated October 21, 2026, which gives affected platforms roughly a month to get their paperwork in order.
What OpenAI actually changed
The update targets what OpenAI calls model-access intermediary platforms. These are the services sitting between developers and the underlying models.
Under the revised policy, that kind of setup now requires explicit sign-off from OpenAI. The requirement applies specifically to platforms that let outside parties reach OpenAI’s models directly.
That distinction matters. The approval rule does not extend to applications that simply use OpenAI models to power their own features.
The line OpenAI is drawing is between using the model and redistributing access to it. One is a customer. The other is closer to a reseller.
Part of a broader policy refresh
The gateway rule did not arrive alone. It is one piece of a wider overhaul of OpenAI’s Usage Policies.
The refresh also clarifies rules on using OpenAI outputs to train other models. Other updated sections cover protections for minors, restrictions on high-stakes uses, and national-security requirements.
The national-security piece sits alongside the new approval requirements in the October 21 revision.
The timing tracks with activity in the intermediary space. Multiple AI gateways and routers have recently rolled out multi-provider routing and their own safety measures.
Multi-provider routing lets a single platform send requests to different AI models depending on cost, speed, or task.
Why gateways became a policy question
When a developer connects to OpenAI directly, OpenAI knows who that developer is. It can enforce its terms on that account.
When access flows through a gateway, the end user is one step removed, and the gateway effectively becomes a middleman with its own customers. That middleman layer makes enforcement harder for any of the issues covered in the broader refresh. Protections for minors, high-stakes use limits, and national-security rules all depend on knowing who is on the other end.
An approval process gives OpenAI a checkpoint. Before a platform can resell access, OpenAI gets to review how that platform operates.
What this means for developers and intermediaries
The most direct impact lands on the gateway and router operators themselves. Any platform offering direct third-party access to OpenAI models now faces a review it did not face before.
For the majority of developers, the practical change may be small. Companies that build products on top of OpenAI models for their own features fall outside the approval requirement entirely.
The developers to watch are those who reach OpenAI models through a third-party gateway. If their gateway does not secure approval before November 20, 2026, their access path could be disrupted.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.