Anthropic has confirmed that it revoked OpenAI’s access to the Claude Code application-programming interface on 29 July, citing a “flagrant violation” of its commercial terms, which prohibit using the service to develop competing artificial-intelligence models. Company spokesperson Christopher Nulty said OpenAI engineers had been employing Claude Code to benchmark and refine GPT-5, the next generation of ChatGPT, contrary to the policy barring use for training rival systems. OpenAI called the move disappointing but said cross-model evaluation is common practice to track progress and improve safety. Hannah Wong, the company’s communications director, noted that Anthropic continues to enjoy access to OpenAI’s own APIs. The dispute surfaces as competition between the two San Francisco-based start-ups intensifies. Anthropic has begun internal testing of its next flagship model, Claude Opus 4.1, with red-teaming expected to last several weeks. Meanwhile, industry speculation points to an August debut for GPT-5, which is expected to deliver higher-quality code generation and improved autonomous-agent capabilities.