Anthropic Blocks OpenAI's Access to Claude AI Models Over Alleged Breach
Anthropic restricts OpenAI from accessing Claude models due to alleged contract breach amid GPT-5 preparations.
Key Points
- • Anthropic blocked OpenAI from accessing Claude models citing breach of contract.
- • OpenAI allegedly used Claude Code for internal testing before GPT-5 launch.
- • Commercial terms forbid reverse engineering and developing competing products.
- • Anthropic continues to allow API access for benchmarking, but OpenAI's eligibility is uncertain.
In a significant move in the competitive landscape of AI technology, Anthropic has blocked OpenAI from accessing its Claude AI models via API, citing a breach of contract related to OpenAI's internal use of Claude's technology ahead of the anticipated launch of GPT-5. The dispute underscores the strategic tensions between the two AI companies, which have been rivals in the rapidly evolving field of AI coding tools.
The allegation arises from claims that OpenAI's technical teams utilized Claude Code for internal testing, which is a violation of commercial terms that expressly prohibit reverse engineering or using Anthropic’s models to develop competing products. Christopher Nulty, a spokesperson for Anthropic, remarked, "it was no surprise to learn OpenAI's own technical staff were also using our coding tools ahead of the launch of GPT-5," highlighting the competitive atmosphere.
Reports indicate that OpenAI integrated Claude into its internal tools for systematic comparisons with its own AI models, focusing on various tasks including coding, creative writing, and testing sensitive content, such as issues related to child sexual abuse material (CSAM) and self-harm. Despite the access blockage, Anthropic has stated it will continue to allow API access for benchmarking and safety evaluations—a standard practice in the industry—though it remains uncertain if OpenAI will qualify for this access in the future.
This incident marks a continuation of Anthropic's trend of restricting OpenAI's access. Notably, they previously limited OpenAI's interaction with the coding tool Windsurf amid acquisition negotiations. Anthropic co-founder Jared Kaplan has emphasized that selling Claude technology to OpenAI could present strategic complications, mentioning concerns over limited computing resources and competitive integrity. He expressed, "I think it would be odd for us to be selling Claude to OpenAI."
As the arms race for AI capabilities escalates, with GPT-5 being developed to outperform competitors in programming tasks, Anthropic's recent actions appear designed to protect its technology from being leveraged in ways that would enhance OpenAI’s competitive position in the market. OpenAI has expressed disappointment over the decision, pointing out their respect for Anthropic’s choice and assuring they have not restricted access to Anthropic APIs.