In a compelling address at CityWeek in London, Randall Kroszner, a senior policy maker at the Bank of England, emphasized the necessity for AI models to be governed by a regulatory ‘constitution’ to mitigate the risks of harmful behavior.
Kroszner, an external member of the Bank of England’s financial policy committee, highlighted the unique challenges posed by AI’s rapid and disruptive innovation compared to more incremental technological advances. “When innovation is incremental, regulators can more easily understand the consequences of their actions and align them with financial stability goals,” he stated. However, the swift and transformative nature of AI presents a far more complex regulatory landscape.
He acknowledged that traditional regulatory tools, like the Bank of England’s Digital Securities Sandbox—which facilitates the use of emerging technologies in the trading and settlement of securities—may not suffice for managing AI’s exponential growth and impact. “Fundamentally disruptive innovations, such as ChatGPT and subsequent AI tools, often scale extraordinarily rapidly, testing the limits of regulatory frameworks,” Kroszner explained. In such scenarios, he suggested, policymakers themselves might need to innovate to keep pace.
Kroszner cited a recent speech by FPC colleague Jon Hall, who warned of the potential risks posed by neural networks evolving into ‘deep trading agents.’ These advanced AI systems could misalign their incentives with those of regulators and the public, potentially amplifying financial shocks and undermining market stability.
To counteract these risks, Kroszner proposed that neural networks should be trained to adhere to a ‘constitution’—a robust set of regulatory rules designed to safeguard financial stability. This constitutional approach could be integrated with sandbox environments to ensure that new innovations are developed responsibly. “In cases where disruptive changes scale too rapidly for sandbox approaches to remain effective, a constitutional framework may be the most suitable solution,” he concluded.
This forward-thinking strategy underscores the importance of proactive and adaptive regulatory measures in the face of AI’s transformative potential, ensuring that innovation can thrive without compromising financial stability.
