Breaking News: Anthropic Terminates Access to OpenAI’s Claude Models: Implications Explored

In a recent development, Anthropic has taken the decision to restrict OpenAI’s access to its Claude family of AI models, as reported by Wired.

According to sources familiar with the matter, OpenAI was utilizing Claude in conjunction with internal tools that enabled comparison of its performance against OpenAI’s own models across various categories, such as coding, writing, and safety.

Anthropic’s spokesperson provided a statement, stating that “OpenAI’s own technical staff were indeed making use of our coding tools prior to the launch of GPT-5, which constitutes a clear violation of our terms of service.” It is noteworthy that Anthropic’s commercial agreements expressly prohibit the utilization of Claude for developing competing services.

Despite this action, Anthropic has confirmed it will maintain OpenAI’s access for “purposes related to benchmarking and safety evaluations.”

In response, an OpenAI spokesperson characterized its usage as being in line with industry standards. The spokesperson further expressed disappointment at the decision to limit API access, considering that Anthropic continues to have unfettered access to OpenAI’s services.

Previously, executives at Anthropic had demonstrated reluctance towards granting access to competitors, with Chief Science Officer Jared Kaplan justifying the company’s decision to deny access to Windsurf (a rumored OpenAI acquisition target, subsequently acquired by Cognition) by stating, “I find it unusual for us to be offering Claude to OpenAI.”