Sign In
By signing in or creating an account, you agree with Associated Broadcasting Company's Terms & Conditions and Privacy Policy.
By signing in or creating an account, you agree with Associated Broadcasting Company's Terms & Conditions and Privacy Policy.
Anthropic has cut off OpenAI access to its Claude AI models for a perceived violation of terms of service, in a major step on the path to the expected release of GPT-5. This suspension follows OpenAI allegedly testing Claude through developer APIs rather than the conventional chat interface to test coding and safety responses. The company is said to have incorporated Claude into its internal tools to compare it to its other models.
An Anthropic spokesperson confirmed the move, saying that OpenAI was using Claude internally in a direct contravention of the terms of use. Although OpenAI is still allowed to use Claude on a limited basis due to safety and benchmarking reasons, it is no longer allowed to use it in a manner that could enable it to develop competitive AI models.
Christopher Nulty, the spokesperson of Anthropic, told Wired that Claude has become a favourite of developers, and the fact that OpenAI plans to use it to complement its own release is not shocking but not allowed. “Regrettably, this is against our terms of service,” Nulty said.
The policy of Anthropic does not allow building or training a competitor AI using its models. Regardless of the prohibition, the company explained that it will still provide API access for benchmarking and safety testing purposes only, which are common in the AI business.
The Chief Communications Officer of OpenAI, Hannah Wong, recognised the cutoff and asserted that it is a standard practice to carry out cross-model assessments. Wong said, “It is an industry practice to test other AI systems to compare the progress and enhance safety.” We understand and appreciate why Anthropic is making this decision, but this is a disappointment given that our API is still available to them.
This is not the only such case. In January, Anthropic also denied access to its models to Windsurf, a coding-orientated AI startup that is said to be acquired by OpenAI. Anthropic Chief Science Officer Jared Kaplan said at the time, “It would be strange for us to sell Claude to OpenAI.”
