Reselling access to OpenAI's models is about to need OpenAI's sign-off. In a post on Wednesday, Oct. 21, 2026, OpenAI said its updated Usage Policies, effective Nov. 20, 2026, require "services that provide direct access to our models such as model routers or AI gateways to get our review and approval before offering that access." Until then, the previous policies stay in effect.
What changes on Nov. 20
OpenAI's Help Center lists four updates:
- Approval requirements for model-access intermediaries.
- Clarified restrictions on using OpenAI models and outputs to train models meant as alternatives to or substitutes for OpenAI's.
- Clarified protections for children and teens and requirements for high-stakes uses.
- National security principles and approval requirements for national security and law enforcement uses.
PANews, summarizing the policy text, reported that apps that simply build OpenAI models into their own features are not covered by the intermediary approval rule, and that the policy also bars extracting protected reasoning content through unauthorized access or reverse engineering.
Partners and SafetyIDs
OpenAI said partners that host or resell its models, including models such as GPT-6 Astra, must keep safeguards "consistent with our own," investigate potential violations and coordinate on enforcement. It is also working with customers to introduce SafetyIDs, which let OpenAI tell individual end users apart during a misuse investigation without learning who they are. OpenAI did not name any router or gateway company in the post.
Why it matters
Gateways are one of the main ways developers switch between models from OpenAI, xAI, Google and others. Related: OpenAI models come to Google Cloud and the compromised LiteLLM gateway release.
