OpenAI is updating its Usage Policies, effective November 20, 2026, so that services offering direct access to its models, including model routers and AI gateways, need OpenAI’s review and approval before they can do so. The rules apply whether developers reach the models directly, through hosting partners, or through reseller partners.

Partners must hold safeguards consistent with OpenAI’s own, investigate potential violations, and coordinate enforcement so that a restriction applied in one place stays effective across the other services serving the same models. OpenAI is also rolling out SafetyIDs, which customers can pass along so OpenAI can tell individual end users apart during an investigation without learning those users’ identities; the stated aim is to act on misuse more precisely with less disruption to everyone else on the account.

The company says enforcement stays tied to documented conduct and risk, and that restrictions get revisited if changes to conduct, commitments or safeguards address the original concern.

Note: OpenAI’s own write-up frames this as aligning safety standards across services that resell access to its models. It does not name any affected provider.

Broad access through consistent standards