Anthropic Removes OpenAI’s Access to Claude Models

Wired has reported that Anthropic has withdrawn OpenAI’s access to its Claude AI models.

Sources cited by Wired indicate that OpenAI was integrating Claude into its internal systems to evaluate its performance in areas such as coding, writing, and safety, making comparisons with its own models.

TechCrunch reached out for comments from both companies. An Anthropic representative informed Wired, “OpenAI’s technical team was utilizing our coding tools prior to the GPT-5 launch,” describing this as “a clear violation of our terms of service.” (Anthropic’s agreements prohibit the use of Claude for developing competing products.)

Nonetheless, Anthropic stated it would continue to allow OpenAI access for “benchmarking and safety evaluations.”

In response, an OpenAI spokesperson remarked, “While we understand Anthropic’s decision to terminate our API access, it’s disappointing, especially since our API remains available to them.”

Anthropic executives have previously expressed hesitance in granting access to competitors. Chief Science Officer Jared Kaplan defended the choice to sever ties with Windsurf (a company reportedly pursued for acquisition by OpenAI, but eventually acquired by Cognition) by stating, “It would be unusual for us to sell Claude to OpenAI.”