- Treasury Secretary Scott Bessent alleged that a Chinese AI model enabled cyberattacks on French refrigeration systems, but offered no public evidence.
- The attacks, attributed by France to Iran's Revolutionary Guard, disrupted roughly 700 supermarkets and a public blood bank, with similar incidents reported across Europe.
- The allegation comes amid a nascent U.S.-China dialogue on AI risk management, underscoring the tension between confrontation and cooperation.
Bessent’s Allegation
Treasury Secretary Scott Bessent on Friday escalated his rhetoric on the dangers of Chinese artificial intelligence, stating that "AI attacks are all coming from Chinese models and they are dangerous," according to a message received Friday. The comment follows his earlier claim that a Chinese AI model was used in cyberattacks on French supermarket refrigeration systems. However, Bessent did not identify the model or its developer, and the allegation has not been independently substantiated in reporting reviewed.
The headline assertion that all AI attacks originate from Chinese models remains unproven. The available evidence supports that a wave of cyberattacks targeted internet-connected refrigeration controllers, causing physical disruption, but the specific role of any AI tool—let alone its origin—is less established. French Foreign Minister Jean-Noël Barrot attributed the attacks to Iran’s Islamic Revolutionary Guard Corps, a finding that addresses the perpetrators but not the technology they may have used.
The Attacks and Their Impact
The cyberattacks, which occurred in October, exploited internet-connected refrigeration controllers used for remote temperature monitoring. Intruders altered settings and triggered prolonged defrost cycles, causing refrigerators and freezers to warm and forcing retailers to discard spoiled food. The disruption affected approximately 700 supermarkets in France and the country’s public blood bank. Similar attacks were reported in Belgium, Ireland, Sweden, Finland, and Estonia. In his earlier remarks on November 12, Bessent referred to 150 grocery stores, a discrepancy that remains unresolved.
The incident highlights the physical consequences of compromised connected systems, extending beyond data theft to operational disruption. Retailers faced spoiled inventory and logistical challenges, while refrigeration and remote-maintenance providers now face scrutiny over the security of internet-accessible control systems. The involvement of a public blood bank brings healthcare infrastructure into focus, though no patient harm has been reported.
Geopolitical and Policy Context
The allegation against Chinese AI models comes amid a delicate balance of confrontation and cooperation between Washington and Beijing. On November 11, Treasury published remarks outlining a new bilateral AI-risk dialogue, including planned exchanges on pre-release model testing and a direct channel for reporting serious AI incidents. That framework, however, describes work underway rather than an operational agreement.
Chinese state-owned newspaper Global Times challenged Bessent’s allegation, arguing that Washington had supplied no public technical evidence and was politicizing cybersecurity. The exchange underscores the broader debate over AI safety and attribution, with policymakers warning of future AI-enabled attacks while grappling with the need for international coordination.
Bessent’s comments also follow his earlier warning that Chinese AI models are “dangerous because they don’t have guardrails.” Yet without public technical evidence identifying the model, its tasks, or whether it materially enhanced the attackers’ capabilities, the AI-specific attribution remains speculative. The strongest supported conclusion is that the refrigeration attacks demonstrate the physical risks of compromised connected systems—while the precise role of a Chinese AI model remains publicly unproven.
Treasury did not respond to a request for comment on the specific evidence behind Bessent’s assertion. The French government has not publicly linked the attacks to any AI tool. As the investigation continues, the focus may shift to securing remote equipment access and maintaining operational resilience, rather than treating cybersecurity solely as a data-protection expense.