OpenAI says model-access partners must maintain aligned safeguards, while updated Usage Policies will require review of direct-access services starting November 20, 2026.
OpenAI published an update on October 21, 2026, describing how it intends to apply safety expectations across services that provide access to its models. The company says it has worked with customers and partners to align policies and is revising its Usage Policies to make requirements clearer for developers.
Under the approach outlined by OpenAI, partners hosting or reselling access to its models are expected to maintain safeguards consistent with OpenAI’s standards, investigate possible violations and take appropriate action. The company also says it will coordinate enforcement so that restrictions remain effective across different services. These expectations apply to access through OpenAI itself as well as through hosting and reseller partners.
OpenAI describes its safeguards as a combination of measures before and after deployment. The company says it evaluates model capabilities and risks, trains models to follow safety requirements and tests safeguards against deliberate attempts to bypass them. After launch, it says it monitors for potential abuse and investigates concerns, including patterns that could indicate harmful use. It also identifies trusted-access programs as an additional layer for some capabilities.
A significant policy change is scheduled to take effect on November 20, 2026. OpenAI says services offering direct access to its models—such as model routers or AI gateways—will need review and approval before providing that access. The announcement does not specify a review timetable or detail how individual approval decisions will be made.
The company is also working with customers on SafetyIDs, which it says could help distinguish individual end users during misuse investigations without giving OpenAI their identities. OpenAI says enforcement decisions will consider documented conduct and risk, with restrictions subject to reassessment when behavior, commitments or safeguards meaningfully change. The update sets out the company’s stated requirements and planned policy changes; the new Usage Policies are due to become effective on November 20.