Anthropic withdrew Openai Access to the API interface to your models on Tuesday, many sources familiar with the case are said by Wired. Opeli was informed that his access was cut off due to violation of the terms of the service.
“Claude Code has become a choice for programmers around the world, so it was not a surprise to find out that OpenAI’s own technical staff also used our coding tools before starting the GPT-5,” said Antropic Christopher Nulty spokesman in a statement for Wired. “Unfortunately, this is a direct violation of our service conditions.”
According to Anthropic Commercial conditions of servicesCustomers cannot operate the service to “build a competitive product or service, including to train competitive AI models” or “reverse engineer or reproduction” of services. This change of OpenAI access to Claude appears as it is chatgptt apparently Preparation for the release of a recent AI, GPT-5 model, i.e. rumors be better in coding.
Opeli connected Claude to its own internal tools using special access to programmers (API), instead of using a regular chat interface, according to sources. This allowed the company to conduct tests to assess Claude’s capabilities in such things as coding and inventive writing in relation to its own AI models and checking how Claude reacted to security hints including categories such as CSAM, self -mutilation and defamation, the sources say. The results facilitate OpenAI compare the behavior of your own models in similar conditions and to make corrections.
“The industry standard is the assessment of other AI systems to compare progress and improve security. Although we respect Anthropica’s decision to cut off our access to the API, it is disappointed, considering that our API remains available to them,” said Hannah Wong, Openai communication director in the statement.
Nulty claims that Anthropic “still will ensure that Opeli has access to API for comparative tests and security, as is the standard practice in the entire industry.” The company did not respond to Wired’s request to explain whether and how the current limitation of API Claude OpenAI would affect this work.
The best technology companies Yanking API Access from competitors has been a tactic in the technology industry for years. Facebook I did the same with Vine belonging to Twitter (which led to allegations of anti -competitive behavior), and in last month Salesforce Limited competitors From access to some data via API Slack. This is not even the first for anthropics. Last month, the company narrow direct access to AI coding direct access to its models after it was remembered that Opennai was to buy it. (This contract has fallen).
Anthropic science director, Jared Kaplan I talked to TechCrunch At that time, withdrawing Windsurf’s access to Claude, saying: “I think it would be strange to sell Claude to OpenAI.”
The day before relieving OpenAI access to API Claude, anthropic announced Recent Claude code speed limits, its tool encoding artificial intelligence, citing Explosive use And in some cases, violation of the terms of the provision of services.