Anthropic has cut off OpenAI’s access to its Claude AI models after it was revealed that OpenAI had used them to compare performance with its own models ahead of the GPT-5 launch. According to Wired, OpenAI had connected Claude to internal tools to evaluate tasks like coding, writing, and safety. Anthropic considers this a violation of its terms of service, which prohibit using Claude to build competing products.
Access will still be allowed for safety and benchmarking purposes. OpenAI called the move disappointing and described its usage as “standard industry practice.” Tensions between the two companies had been growing, as Anthropic had previously restricted access to other competitors as well.