xAI Reportedly Trained Its Coding Models on Claude Outputs for Months Before Being Cut Off

Elon Musk xAI trained its coding-focused AI models on outputs generated by Anthropic Claude for an extended period before Anthropic detected the pattern and cut off access. The report claims xAI engineers systematically queried Claude API endpoints to generate training data for xAI own code generation models.

The alleged practice falls into a gray area of AI development: using one model outputs to train another, sometimes called model distillation or synthetic data generation. While technically feasible through API access, most AI providers terms of service prohibit using their models to train competing systems.

Anthropic reportedly identified unusual query patterns consistent with training-data harvesting and terminated the associated accounts. Neither xAI nor Anthropic have publicly commented on the specific allegations. The incident highlights the tension between open API access and competitive dynamics in the AI industry.