Anthropic revoked OpenAI’s access to its Claude API after discovering that OpenAI engineers were using Claude Code to test GPT-5. The move, announced on August 1, stems from a violation of Anthropic’s terms of service.
“Claude Code has become the go-to choice for coders everywhere, and so it was no surprise to learn OpenAI’s own technical staff were also using our coding tools ahead of the launch of the latest version of ChatGPT in GPT-5,” Anthropic spokesperson Christopher Nulty told Wired. “Unfortunately, this is a direct violation of our terms of service.”
Testing, not training: OpenAI defends API use
Training competing AI models is directly against Anthropic’s commercial terms of service, which states users cannot employ Claude to “access the Services to build a competing product or service, including to train competing AI models except as expressly approved by Anthropic; reverse engineer or duplicate the Services.”
OpenAI said it used the API to benchmark progress, not to work on its own code.
“It’s industry standard to evaluate other AI systems to benchmark progress and improve safety,” OpenAI Chief Communications Officer Hannah Wong told Wired.
Using the API instead of Claude’s external chat window lets OpenAI run tests against GPT-5.
Anthropic said it would allow OpenAI to have API access for benchmarking and safety evaluations specifically.
This is not the first time the Claude maker has acted wary of one of its major rivals getting too close a look at its models. Anthropic closed off API access to Windsurf earlier this year after rumors suggested OpenAI might buy the coding tool firm.
GPT-5 is expected to come out in August
GPT-5 is slated for release in August; it is expected to be a mixture of the types of models OpenAI offers, including reasoning models such as o3. The free upgrade will choose which model to use based on the query, as opposed to offering the user a drop-down menu.
CEO Sam Altman teased GPT-5 on Sunday, showing its answers to a request for a thought-provoking television series about AI.
Looking for a different advanced AI? Google Gemini with Deep Think, which offers several reasoning agents working in parallel, is out for subscribers to the company’s AI Ultra services.
