The top AI labs are all racing towards the same goal of AGI, but there seems to be no love lost between them.
Anthropic has cut off OpenAI’s access to its model APIs. Anthropic informed OpenAI that it had revoked access to its models for violating its terms of service. “Claude Code has become the go-to choice for coders everywhere, and so it was no surprise to learn OpenAI’s own technical staff were also using our coding tools ahead of the launch of GPT-5,” Anthropic spokesperson Christopher Nulty told Wired. “Unfortunately, this is a direct violation of our terms of service,” he added.
As per Anthropic’s commercial terms of service, customers are barred from using the service to “build a competing product or service, including to train competing AI models” or “reverse engineer or duplicate” the services.
OpenAI said that it was “disappointing” that their access had been cut off. “It’s industry standard to evaluate other AI systems to benchmark progress and improve safety. While we respect Anthropic’s decision to cut off our API access, it’s disappointing considering our API remains available to them,” OpenAI’s chief communications officer Hannah Wong said.
Anthropic’s revocation of its API for OpenAI employees comes at a time when OpenAI is preparing for two big releases — its flagship GPT 5 model, as well as a new open-source model. The open-source model is already significantly delayed, and there’s breathless anticipation for GPT 5. GPT 4.5 had been underwhelming, and users are expecting OpenAI to come up with a really powerful model in GPT 5, and again establish a lead on the benchmarks, which now see Google and xAI’s Grok at the top.
This isn’t the first time that Anthropic has cut off access to its models to OpenAI. When it had been rumoured that Windsurf had been acquired by OpenAI, Anthropic had promptly cut off access to its models to Windsurf. Anthropic clearly felt that letting its models to continue to be hosted on a service owned by OpenAI would likely give its rival too much data about how its models were being used. Anthropic’s coding models are especially popular, and it wants to closely guard its trade secrets against OpenAI. However, when OpenAI’s deal with Windsurf had fallen through, Anthropic had restored access to its models for Windsurf.
Anthropic has now revoked access to its API for all of OpenAI. It’s unclear what prompted it to take the step — it would’ve been normal for OpenAI to benchmark its models against competitors and to work on safety issues, but Anthropic has said that OpenAI had violated its terms of service, which could indicate that the company was using its models to improve its own. What makes this entire situation even more interesting is that Anthropic’s founders were formerly with OpenAI, and had broken away from the company and started their own because they claimed that they weren’t satisfied with OpenAI’s stance on security. It remains to be seen how this situation resolves — as things stand, Anthropic has access to OpenAI’s models, while OpenAI does not have access to Anthropic’s — but as the stakes in the AI race get higher, companies seem to be getting ever more careful about maintaining their competitive advantages.