World financial leaders and policymakers gathered in Bangkok last week to address growing concerns about the risks posed by cyber-attacks amplified by artificial intelligence (AI). The discussions highlighted the increasing threats to financial stability and consumer trust in an era of rapid technological advancement.

Kristalina Georgieva, managing director of the International Monetary Fund (IMF), cited a recent incident in France where hackers manipulated supermarket refrigerators to raise temperatures to 50°C, causing widespread spoilage and significant losses. This example underscored the vulnerabilities of interconnected systems and the escalating costs of cybercrime. Georgieva emphasized that AI is magnifying the potential scale and sophistication of digital fraud, comparing it to “digital fraud on steroids.”

While some analysts have chiefly worried about AI’s impact on financial markets, such as the possibility of a sharp correction in share prices, the IMF chief highlighted a different but related danger: the erosion of public trust in the financial system. AI-driven cybercrime, including convincing fake endorsements on social media and automated email scams, could deter consumers from using mainstream banks. Reflecting on her upbringing in Bulgaria under communist rule, Georgieva warned that a loss of confidence might prompt many people to withdraw cash and avoid financial institutions altogether, likening this to putting money in "shoeboxes under their beds."

Concerns over AI’s role in financial services were echoed by Jamie Dimon, chief executive of JPMorgan Chase, who announced the bank’s refusal to allow AI agents to manage customer accounts. He warned of the risks posed by unregulated “rogue agents,” which could number in the thousands, complicating liability issues if AI-driven decisions lead to financial losses or illegal activities. Regulators fear that widespread use of AI financial agents, such as Meta’s Muse, which automatically moves funds to optimize savings rates, might accelerate bank runs by facilitating rapid withdrawals in uncertain times.

Andrew Bailey, governor of the Bank of England and chair of the Financial Stability Board, highlighted another emerging challenge: the potential for AI models to “self-replicate” within domestic internet-connected devices, raising questions about responsibility and liability if such devices malfunction or cause damage. Some French supermarkets have temporarily disconnected their smart refrigerators as a precaution against hacks, illustrating the growing trend of reverting to “dumb” appliances to avoid cyber risks.

Bailey stressed the fundamental importance of trust in money for societal functioning, noting that without it, economic operations falter. The convergence of AI, cybercrime, and the financial system now extends beyond traditional banking to encompass social media, payment applications, and other digital financial tools, prompting calls for urgent, coordinated action.

Despite the urgency, geopolitical tensions and divergent priorities complicate international cooperation. In the United States, the administration has adopted the term “superintelligence” for AI during ongoing debates, while efforts to establish regulatory frameworks remain nascent. A taskforce chaired by the director of national intelligence has begun to develop mechanisms for reporting AI-related breaches but has yet to implement mandatory pre-launch testing for AI models.

Looking ahead, the United Kingdom plans to prioritize AI safety as part of its upcoming G20 presidency, and G7 finance ministers have committed to simulations testing the resilience of the global financial system against cyber-attacks. However, experts caution that without effective global coordination and legal clarity, misuse of AI could trigger financial instability, undermining the foundational trust necessary for economic stability and posing risks that transcend traditional regulatory frameworks.