Anthropic cuts off OpenAI’s access to its Claude models

In a significant move, Anthropic has officially cut off OpenAI's access to its Claude AI model suite, as reported by Wired. Sources revealed that OpenAI was utilizing Claude to benchmark its own AI capabilities against those of Anthropic across various domains, including coding, writing, and safety assessments. An Anthropic spokesperson confirmed to Wired that OpenAI's technical team had been using their coding tools in the lead-up to the anticipated launch of GPT-5. This action was deemed a breach of Anthropic’s terms of service, which explicitly prohibit the use of Claude for developing competing services. Despite the termination of access, Anthropic stated it would still permit OpenAI to use its models for benchmarking and safety evaluations. In response to the news, an OpenAI representative expressed disappointment over the decision, emphasizing that while they respect Anthropic's choice, their own API remains accessible for Anthropic’s use. This development underscores Anthropic's growing caution regarding providing access to its technology for competitive purposes. Previously, Chief Science Officer Jared Kaplan defended the company’s stance in relation to another competitor, Windsurf, articulating that it would be unconventional for them to sell Claude to a rival like OpenAI.