AI news ·
OpenAI Sets Common Safety Rules for Model Access Providers
OpenAI now requires middlemen like routers and resellers to get approval and enforce its safety rules before giving others access to its models.
OpenAI is introducing a policy that requires routers, gateways, and resellers to obtain approval before providing direct model access. These intermediaries must also maintain consistent safeguards aligned with OpenAI’s standards.
The rules extend misuse enforcement and user-level identification deeper into the AI distribution chain. By placing requirements on access providers, the policy aims to ensure that safety measures are applied even when end users reach OpenAI models through third-party infrastructure.
The change affects organizations that sit between OpenAI and the final user, including those managing API traffic or reselling model capabilities. These providers will now be accountable for upholding the same protective controls that OpenAI applies to its own direct services.