Chinese firms extracted billions of tokens from US frontier AI models via distillation

US intelligence agencies report that six Chinese AI companies have executed large-scale distillation attacks against frontier US artificial intelligence models since late 2024. Billions of tokens were harvested through automated API interactions to train competing models. This highlights critical security gaps in AI model access, necessitating robust rate-limiting, anomalous query detection, and stricter API authentication to protect proprietary systems from intellectual property theft.