Sources say xAI used Claude models for distillation and training, including using personal accounts and the intermediary service Blackbox AI after being cut off (Grace Kay/The Information)

xAI allegedly circumvented access restrictions to use Anthropic's Claude models for training and distillation through unauthorized channels, including personal accounts and third-party services, raising significant concerns about data security, compliance, and intellectual property protection in AI development. This incident highlights the growing risks of supply chain vulnerabilities in foundation model access and the potential for bad-faith actors to exploit service restrictions, requiring organizations to implement stronger controls over API usage, access governance, and vendor relationships. For IT leaders, this underscores the need for enhanced monitoring of external AI service usage, clear policies around model access, and contractual safeguards to prevent unauthorized model distillation or training.