Anthropic Blocks OpenAI Access to Claude Models Over Alleged Breach

Anthropic has blocked OpenAI’s access to its Claude artificial intelligence models, alleging a breach of contract. The move comes as OpenAI prepares for the release of its highly anticipated GPT-5 model.

According to reports from WIRED, Anthropic claims that OpenAI’s technical teams utilized Claude’s coding tools for internal testing in the lead-up to GPT-5’s launch. Anthropic’s commercial terms of use explicitly prohibit the reverse engineering of its models and their use in the development of competing products.

Christopher Nulty, a spokesperson for Anthropic, told WIRED, “it was no surprise to learn OpenAI’s own technical staff were also using our coding tools ahead of the launch of GPT-5.” The report detailed that OpenAI had integrated Claude into its internal development tools via standard API access, systematically comparing Claude’s performance against its own AI models. These tests reportedly covered a wide range of tasks, from coding and creative writing to the evaluation of security-sensitive content, including child sexual abuse material (CSAM), self-harm, and defamation.

Anthropic stated that it would continue to permit API access “for the purposes of benchmarking and safety evaluations,” describing this as a standard industry practice. However, the company did not clarify whether OpenAI would still qualify for this limited form of access moving forward.

Responding to the action, OpenAI spokesperson Hannah Wong described the decision as “disappointing” but affirmed the company’s respect for Anthropic’s choice. She also highlighted that OpenAI had not, in turn, restricted Anthropic’s API access.

This is not the first instance of Anthropic cutting off access in response to OpenAI’s activities. In June, Anthropic significantly restricted Claude’s access for the coding tool Windsurf after reports emerged that OpenAI was considering acquiring the company. At an industry conference, Anthropic co-founder Jared Kaplan cited both limited compute resources and strategic concerns, stating, “I think it would be odd for us to be selling Claude to OpenAI.” While that acquisition did not materialize, with Windsurf’s executives and key talent ultimately joining Google and the startup merging with code software provider Cognition, OpenAI no longer has access to Anthropic’s API through that avenue either.

The escalating competition in AI code generation appears to be a significant factor driving Anthropic’s recent decisions. Internal reports suggest that GPT-5 is aiming to surpass rivals in programming tasks, an area where Claude Code has historically held a strong lead. Anthropic’s actions seem strategically designed to prevent its proprietary Claude data from being incorporated into or used to enhance competing models.