OpenAI requires approval for third-party model routers and AI gateways
OpenAI’s updated usage policy takes effect on November 20, 2026. Model routers and AI gateways must receive OpenAI’s review and approval before providing third parties with access to OpenAI models. Services that use OpenAI models only within their own applications are exempt from this requirement. Partners must maintain security safeguards consistent with OpenAI’s standards and investigate potential violations. The policy also urges partners and the industry to establish mandatory reporting mechanisms for AI security incidents. Separately, it prohibits using OpenAI models or their outputs to prepare, filter or provide training data or signals for competing models, including constructing training tasks, reinforcement-learning environments, or scoring and ranking feedback.
The policy directly affects API intermediaries and model gateways that provide third-party access. It also sets safety, violation-investigation and anti-distillation requirements for approved partners.
OpenAI开始给中转站发“经营牌照”,野生中转站们的好日子到头了。
日前, OpenAI发布公告,自11月20日起新版使用规则生效:模型路由、AI网关这类向第三方提供模型直连的服务须先经OpenAI审查批准,自家App内调用不受影响。
这意味着,以转接调用 OpenAI 大模型为主要商业模式的中转站面临着严峻考验。
其实很早之前就有专业团队通过实测发现,同样是OpenAI的API,进入中转站以后会出现“降智”的情况,中转站API的跑分,要比OpenAI官方API最多低将近一半。
不过OpenAI严格把控API,并不只是为了保证API的质量和用户体验,更重要的一点是,中转站们抢了OpenAI的生意。
因为中转站的上游来源并非只有OpenAI官方直营,有的可能来自于“低价账号池”。基于此,中转站就可以用超低价格出售API,进而影响到了OpenAI自身的收入。