Anthropic revokes Claude's access to OpenAI.

OpenAI can no longer use Anthropic 's Claude models. The San Francisco startup has revoked access to its API for violating its terms of service. Engineers at the company led by Sam Altman allegedly used Claude Code during the development of GPT-5 .

Clash of AI startups

Competition among companies developing AI models is quite fierce. Meta has offered millions of dollars to leading experts in the field, including those from OpenAI. The latter, according to Anthropic, instead wants to "steal" Claude's capabilities to improve GPT-5 .

API access is typically used to evaluate performance through benchmarking. The terms of service prohibit using Claude to develop competing products. Anthropic has revoked the API for access to Claude Code , an AI agent for writing code. Spokesperson Christopher Nulty stated:

Claude Code has become the go-to tool for programmers around the world, so it was no surprise to discover that OpenAI's engineering staff was also using our programming tools before the launch of GPT-5. Unfortunately, this is a direct violation of our terms of service.

Hannah Wong, OpenAI's communications director, responded:

It's industry practice to evaluate other AI systems to test progress and improve security. While we respect Anthropic's decision to discontinue access to our API, we're disappointed that our API remains available to them.

Nulty emphasized that OpenAI will still be able to use API access for benchmarking and security evaluation, but did not clarify how. At the end of July, Anthropic announced new weekly limits (starting August 28) on Claude Code access after detecting excessive use, including violations of the terms of service.