US Treasury Secretary Scott Bessent has linked AI-driven cyberattacks to Chinese artificial intelligence models. He specifically cited an incident between November 12 and 13, 2026, where a Chinese model allegedly disabled refrigeration systems in 150 to 700 French grocery stores, attributing the operation to Iranian operatives using the technology.

Bessent has also raised concerns regarding industrial-scale distillation, where smaller models are trained on outputs from more advanced ones. He suggested that Chinese models, such as Moonshot AI’s Kimi K3, might contain watermarks indicating they were trained on US-developed models, specifically mentioning Anthropic’s Fable model.

In September 2026, US agencies accused six Chinese firms of conducting distillation attacks on American models, an activity reportedly dating back to late 2024. While no formal sanctions have been imposed, Bessent indicated that such evidence could lead to the inclusion of these companies on the US trade blacklist or the implementation of sanctions.

Despite the accusations, the claims regarding the French grocery store attack remain unverified. Meanwhile, US officials have proposed establishing a hotline with China to manage risks related to AI agents and cyberattacks, though the effectiveness of this diplomatic effort remains uncertain amidst the current geopolitical climate.