OpenAI's updated usage rules require model routers and AI gateways offering direct access to its models to obtain review and approval. The requirement takes effect on 20 November. The company also describes shared safeguards and coordinated responses to misuse across hosting and resale partners, with user identifiers intended to support investigations without revealing a person's identity to OpenAI.
OpenAI’s updated usage policies require review and approval for services that offer third parties direct access to its models. Model routers and AI gateways, which connect users to model services, fall within the announcement. The requirement takes effect on 20 November. It adds an approval step for intermediaries alongside the company’s work with partners on consistent safeguards.[1], [2]
The company says partners hosting its models and resellers must maintain compatible safeguards, investigate possible violations and act when concerns are identified. Coordination is intended to keep access restrictions effective across services. OpenAI describes both protections built into models and the monitoring systems and human processes around them, while acknowledging differences in how partner services operate.[1]
OpenAI identifies compromised accounts, leaked API keys and end-user misuse as problems customers can encounter. SafetyIDs, identifiers being introduced with customers, are intended to distinguish end users during investigations without revealing their identities to OpenAI. The company says restrictions can be reconsidered after meaningful changes in conduct, commitments or safeguards address the concern. These remain announced policies and mechanisms rather than an independent audit of their effectiveness.[1]
