In a bold move shaking up the AI world, Anthropic has blocked OpenAI from accessing its Claude AI models. The decision follows claims that OpenAI violated Anthropic’s usage terms by using Claude’s tools to evaluate and potentially enhance its own upcoming model, GPT-5.
Why Anthropic Has Blocked OpenAI’s Access
Sources told Wired that OpenAI had been using Claude for internal testing — comparing its performance in:
- Coding
- Writing tasks
- Safety evaluations
Anthropic claims this kind of usage crossed the line, especially because their commercial agreement clearly bans using Claude to build competing tools.
“OpenAI’s own engineers were using our tools while preparing for GPT-5,” Anthropic stated.
“That directly violates our terms of service.”
OpenAI Responds After Anthropic Has Blocked Access
While Anthropic has blocked OpenAI, the latter isn’t backing down quietly. OpenAI insists its use of Claude was routine and accepted across the AI industry.
A spokesperson said:
“We respect Anthropic’s decision, but we’re disappointed — especially since they still have access to our API.”
This reflects the growing friction between collaborating and competing AI companies.
Limited Access Will Still Continue
Although Anthropic has blocked OpenAI’s full API access, it’s not a complete shutdown.
The company confirmed that OpenAI can still use Claude in a limited capacity, specifically for:
- Benchmarking performance
- Conducting safety-related research
This suggests a desire to maintain some level of cooperation, even as competition heats up.
Not the First Time Anthropic Has Blocked a Competitor
This isn’t the first instance where Anthropic has blocked access to rivals.
Earlier this year, it shut down access to Windsurf, a small AI company rumored to be on OpenAI’s radar. Windsurf was later acquired by Cognition.
Anthropic’s Chief Science Officer, Jared Kaplan, summed it up:
“It would be strange to sell Claude access to a direct competitor like OpenAI.”
Key Takeaways
- Anthropic has blocked OpenAI from using its Claude AI models due to a terms-of-service violation.
- OpenAI reportedly used Claude’s tools to help prepare GPT-5, which Anthropic says is not allowed.
- OpenAI called the usage standard practice, but said it respects the decision.
- Limited access will continue for safety and benchmarking purposes.
- This move highlights growing competition and stricter boundaries between top AI companies.