The IMF's spring meetings in Bangkok produced an unlikely mascot for systemic financial risk: French supermarket fridges hacked to 50°C, spoiling food en masse. IMF managing director Kristalina Georgieva used the episode to illustrate how AI is 'putting the potential of digital fraud on steroids' — and how the attack surface now extends far beyond banks into every internet-connected device. The deeper concern isn't rogue appliances. It's the compounding collision of three forces: AI's ability to generate scams at industrial scale, the spread of autonomous AI financial agents that move money without human oversight, and a regulatory architecture built for a world where humans made the decisions. JP Morgan CEO Jamie Dimon warned his bank would block AI agents acting on behalf of customers, flagging the risk of 'thousands of rogue agents.' Meta's Muse is already offering to scout savings rates and shift user deposits automatically. Financial regulators see two destabilising vectors in AI agents. First, herding: if millions of AI-driven deposits chase the same best rate simultaneously, the resulting capital flows could accelerate bank runs in moments of uncertainty. Second, liability ambiguity — when an AI agent makes a disastrous or criminal financial decision on your behalf, existing law cannot assign blame. Bank of England governor Andrew Bailey, who chairs the Financial Stability Board, pointed to Anthropic's warning that AI models might soon self-replicate in domestic devices, asking who bears the cost when an internet-enabled appliance is hijacked. The 2008 crisis demonstrated what happens when liability structures become so complex they're impossible to value — panic fills the gap. AI agents create an analogous fog, only faster. Case law will eventually clarify responsibility, but that clarity arrives only after something goes badly wrong. Georgieva invoked her Bulgarian childhood behind the iron curtain, where people kept cash in shoeboxes because they didn't trust institutions. She warned the same dynamic could return if consumers lose faith in digital finance. International response is moving, but unevenly. G7 finance ministers announced cyber-resilience simulations for the global financial system. The UK pledged to prioritise AI safety during its G20 presidency next year. Trump's superintelligence taskforce, chaired by intelligence director Jay Clayton, is setting up breach-reporting mechanisms for AI labs — though notably not requiring pre-launch model testing. Fed chair Kevin Warsh dutifully used Trump's mandated 'superintelligence' rebrand on stage in Bangkok. The timing is poor. The country driving the AI boom is governed by an administration more interested in renaming the technology than regulating it. The questions at stake — who is liable when autonomous agents fail, how to prevent AI-accelerated bank runs, whether consumers will trust systems they can't understand — demand coordinated international action at a moment when the appetite for cooperation is at a low ebb. Bailey's summary was blunt: 'If we don't have trust in money, we don't operate as a society.' The irony is that the rational consumer response — rejecting connected devices, pulling money from digital systems — would itself damage the economy. The system needs trust to function, and trust is precisely what AI-enabled fraud, liability ambiguity, and autonomous agents are eroding.
