(19659001) All products featured on WIRED were independently selected by our editors. We may receive compensation if you purchase products via these links.
Anthropic has revoked OpenAI ‘s API access to their models on Tuesday. OpenAI was informed of the termination of its API access due to a violation of the terms and conditions.
Anthropic spokesperson Christopher Nulty told WIRED that “Claude Code is the go-to tool for coders around the world, so it’s no surprise that OpenAI’s technical staff was also using our coding software ahead of the GPT-5 launch.” “Unfortunately, it is a violation of our terms.”
Anthropic’scommercial terms of services prohibit customers from using the service “to build a competing product or a service, including to teach competing AI models”or “reverse-engineer or duplicate” the service. OpenAI’s new access to Claude is a result of the ChatGPT maker preparing a new AI model called GPT-5 that is rumored to be more coding-friendly. Sources claim that OpenAI
was using APIs to integrate Claude with its internal tools instead of the chat interface. The company was able to run tests on Claude to compare his abilities in areas like coding, creative writing, and self-harm against its AI models. They could also check how Claude responded when given safety-related prompts, such as CSAM. The results allow OpenAI to compare its own models under similar conditions, and make necessary adjustments.
It’s standard practice to evaluate other AI systems in order to benchmark progress and improve security. Hannah Wong, OpenAI’s chief communication officer, said that while we respect Anthropic for their decision to block our API access, this is disappointing because our API is still available to them.
Nulty states that Anthropic “will continue to ensure OpenAI has API accessibility for the purposes for benchmarking and safety assessments as is standard across the industry.” However, the company did not respond WIRED’s question about how OpenAI’s Claude API restriction will impact this work. Since years, top tech companies have been yanking APIs from their competitors. Facebook did the exact same thing to Twitter-owned Vine (which led to accusations of anticompetitive behaviour) and Salesforce restricted competitors’ access to certain data via the Slack API last month. Anthropic has done this before. Last month, Anthropic restricted Windsurf’s access to its models because it was rumored that OpenAI was going to buy it. This deal did not go through ). Anthropic’s chief scientist officer Jared Kaplan spoke to TechCrunch at the time
about revoking Windsurf’s accessto Claude. “I think that it would be odd for me to sell Claude to OpenAI.”
