Bank of England governor Andrew Bailey today warned that advanced artificial intelligence poses risks to financial infrastructure.
Bailey outlined his concerns in a letter to G20 finance ministers and central bank governors. The document was made public shortly before the group convened in Asheville, North Carolina for its annual meeting. Bailey penned the document in his capacity as the chair of the Financial Stability Board, an international body that oversees the health of the financial system.
The letter warns that financial markets are vulnerable to a correction because of multiple “fragilities.” Bailey listed “stretched asset valuations” in the AI market as one such fragility. He also pointed to the role of debt in circular investment deals between AI companies and hyperscalers.
The rest of the two-page document mostly focuses on large language models’ ability to launch cyberattacks.
According to Bailey, one reason frontier LLMs pose a risk is that many important financial systems are interconnected. The world’s top investment firms complete numerous transactions with one another each day. Furthermore, they often do so using technology from the same third-party providers. A breach at such a key provider can potentially compromise multiple financial institutions.
The letter goes on to make several cybersecurity suggestions for the industry. Bailey recommends that decision-makers prepare for disruptions that affect multiple financial institutions or technology suppliers at once. He also calls for the implementation of workflows that make it possible to “restore critical systems and data from ‘bare metal’ following a significant cyber incident.” Bare metal servers are machines that lack an operating system.
The letter notes that AI models’ ability to rapidly spot software vulnerabilities will require banks to roll out patches faster. That’s challenging for multiple reasons. When a new exploit is discovered, engineers must not develop a patch but also test that it works as expected before releasing it. The latter task historically involved a significant amount of manual work.
“Frontier AI offers significant opportunities to strengthen cyber defence; but recent developments highlight the importance of ensuring that advances in capability are matched by resilience and preparedness,” Bailey wrote.
Last week, more than 100 banks, tech firms and other companies published a public letter warning about the cybersecurity risks of AI. The signatories believe that the frequency of LLM-driven hacking campaigns will increase significantly in the coming months. The letter urges organizations to bolster their defenses by providing their technology teams with AI cybersecurity tools.