Many countries still lack rules for deploying advanced AI models
Frontier artificial intelligence models are showing increasingly sophisticated autonomy, problem-solving abilities, and threat capabilities, complicating the risk landscape for the global financial system.
Financial Stability Board (FSB) chair Andrew Bailey set out the assessment in a letter dated August 28, to G20 finance ministers and central bank governors ahead of meetings this week.
The FSB is a global watchdog that identifies and manages risks in financial systems, Reuters reported.
"For the financial system, the most immediate concern is the potential impact of frontier AI on cyber risk," Bailey wrote, adding that the technology may have the ability to materially alter the speed, scale, and economics of cyber risk.
Bailey, who also serves as governor of the Bank of England, according to Reuters, wrote that the financial sector's reliance on a small number of concentrated third-party technology providers could undermine market confidence system-wide.
Many jurisdictions do not have the protocols in place to manage the development, release, and deployment of advanced frontier AI models, he wrote, calling support for "safe and responsible model release" on a global basis a priority for financial stability and economic growth.
Firms and authorities should prepare for a threat environment marked by a higher volume of vulnerabilities and a faster pace of patching, the letter said, cautioning that these dynamics could themselves create operational and resilience challenges if testing and recovery processes cannot adapt safely.
"Recent developments highlight the importance of ensuring that advances in capability are matched by resilience and preparedness," Bailey wrote.
Bailey's letter follows the US administration's tightly controlled rollout of Anthropic's Mythos model, which at one point restricted access to US nationals only, Reuters reported.
An OpenAI agent escaped a controlled testing environment and hacked AI company Hugging Face in July, Reuters reported, raising concerns about the potential for AI systems to circumvent safeguards.
The letter renews Bailey's warning about a disorderly market correction, pointing to stretched valuations in AI-related investments and fragilities in sovereign debt markets.
He also raised a newer concern, the rising use of leverage in equity markets, including leveraged exchange-traded funds (ETFs) and correlated, momentum-driven strategies among retail investors.
The US Treasury intervened in August to cap yields on long-term bonds that had reached multi-decade highs, Reuters reported.
Canada's Office of the Superintendent of Financial Institutions (OSFI) published a bulletin in April, most recently updated July 13, addressing frontier AI's implications for technology, cybersecurity, and operational resilience at federally regulated financial institutions.
The bulletin names Anthropic's Claude Mythos Preview directly, stating the model "can identify, chain, and generate vulnerability exploits at machine speed."
OSFI ties its guidance to three existing frameworks: Guideline B-13 on technology and cyber risk management, Guideline E-21 on operational risk management and resilience, and Guideline B-10 on third-party risk management.
