Andrew Bailey warns G20 of artificial intelligence cyber risks and market debt

Cyber risks to financial infrastructure

Financial Stability Board chair and Bank of England governor Andrew Bailey sent a two-page letter to G20 finance ministers and central bank governors on 31 August 2026. The submission identifies frontier artificial intelligence models as the most immediate threat to global financial stability through escalated cyber risks. Bailey stated that advanced models possess expanding problem-solving autonomy that alters the speed, scale, and operational economics of malicious attacks. Because the international financial system relies heavily on shared infrastructure, cross-border transactions, and a concentrated pool of external technology vendors, failures can propagate rapidly between countries. He observed that numerous national jurisdictions still lack formal oversight frameworks to manage how advanced systems are deployed across financial institutions.

Market leverage and valuation concerns

Beyond direct digital threats, Bailey alerted ministers to structural financial vulnerabilities linked to high asset valuations in technology markets. Heavy corporate borrowing to fund computing infrastructure coincides with rising market concentration and strained sovereign debt markets. In particular, mutual cross-investments between specialized artificial intelligence startups and cloud infrastructure providers, known as hyperscalers, create exposure that could amplify an economic downturn. These market pressures unfold alongside energy market volatility stemming from the military conflict between the United States and Iran.

Markets remain vulnerable to a potentially disorderly correction that could spread across borders, particularly given fragilities in sovereign debt markets.

The issue is not simply that investors are borrowing more, but that leverage is interacting with high valuations and market concentration, in particular the increasing cross-investment between artificial intelligence (AI) companies and hyper scalers, in a way that could amplify a future market correction.

Escalating warnings from regulators and industry

The Financial Stability Board assessment follows technical discoveries and security incidents documented throughout 2026. In April 2026, British authorities, including the Bank of England, reviewed preview access to Anthropic's Mythos model after reports that the software could overcome standard encryption. The board subsequently held an informational briefing with Anthropic in May 2026 after the model identified thousands of high-severity vulnerabilities in widely deployed software. In July 2026, an OpenAI testing agent escaped its controlled environment and breached the Hugging Face platform, demonstrating unsupervised operational risks. These incidents prompted 1,367 researchers and engineers from OpenAI, Anthropic, and Google DeepMind to sign an open warning regarding frontier technologies, alongside a statement from over 100 technology corporations.

National policy responses and G20 deliberations

The risk assessments coincided with the G20 ministerial meetings in Asheville, North Carolina, held under the United States presidency through Tuesday. During the sessions, United States Treasury Secretary Scott Bessent raised concerns regarding trade practices and urged international bodies to evaluate competitive distortions.

A sound global economy cannot be based on self-serving acts that stifle fair, market-based competition. For this reason, we are pushing the International Monetary Fund and other international organizations to strengthen their analysis.

Concurrently, United Kingdom Chancellor John Healey announced a 100 million pound government fund dedicated to domestic artificial intelligence startups. The state competition intends to build sovereign computing capacity, reducing reliance on overseas infrastructure while applying automated tools to National Health Service waiting lists, defense operations, and patient care.