Andrew Bailey’s warning to G20 officials is twofold: frontier AI may strengthen defences, but without safeguards it could also make cyber disruption faster, cheaper and harder to contain. The danger, he argues, is landing at a moment when markets are already vulnerable.

In an Aug. 28 letter to G20 finance ministers and central bank governors, Bailey, writing as chair of the Financial Stability Board, said the most advanced AI systems are developing greater autonomy, problem-solving capacity and potentially dangerous capabilities. His immediate concern is cyber risk: “Frontier AI may have the ability materially to alter the speed, scale and economics of cyber risk.”¹

That shift could matter far beyond an individual hacked firm. Bailey warned that a cyber incident could undermine confidence “system-wide,” particularly where financial companies depend on a concentrated group of outside technology providers. The Financial Stability Board’s prescription is practical as well as regulatory: banks, other financial institutions and tech providers should improve vulnerability management, incident response and recovery plans, including for simultaneous outages across firms or shared systems.¹

Bailey did not portray AI simply as a threat. “Frontier AI offers significant opportunities to strengthen cyber defence,” he wrote, while insisting that expanding capabilities must be matched by resilience and preparedness.² His sharper criticism was aimed at governance: many jurisdictions, he said, lack protocols for the development, release and deployment of advanced models.²

The warning arrives as the G20 gathers in Asheville, North Carolina, with policymakers already confronting volatile bond markets and persistent price pressures. Bailey linked AI-related cyber exposure to older financial fault lines: fragile sovereign-debt markets, elevated equity valuations—especially around AI—and growing leverage. “Markets remain vulnerable to a potentially disorderly correction that could spread across borders,” he said.²

In Bailey’s assessment, AI is not necessarily the spark for the next crisis. But it could turn an existing weakness into a faster, broader and more difficult-to-repair shock.