How artificial intelligence could upend global financial stability

Central bankers on both sides of the Atlantic are paying attention to the ways generative artificial intelligence may create hidden new risks to the global financial system.

Why it matters: As banks, hedge funds and other major institutions explore using AI in their trading strategies, it could fuel rapid flows of capital and create hidden risks of financial crisis. That, at least, is the fear of some who study financial risk.

• Imagine a future in which big-money investors worldwide rely on AI-driven algorithms to identify and execute trades. It is plausible that trillions of dollars could move in and out of assets rapidly.
• That could have massive economic consequences that regulators, and even the humans running those financial entities themselves, don't understand.

State of play: In a recent paper, Jon Danielsson of the London School of Economics and Andreas Uthemann of the Bank of Canada sketch out how AI may prove highly useful for officials regulating the behavior of individual firms.

• "AI excels and outperforms humans in risk modelling and management," they write, and it "is making rapid inroads into detecting fraud, consumer protection" and more.
• But, they argue, things get considerably more problematic when you go from the micro level to the macro.

What they're saying: "AI systems used by the private sector are better at finding optimal solutions than their human counterparts, but at the risk of such solutions being socially undesirable," the authors wrote.

• "Financial markets are particularly susceptible to such outcomes because they have strong complementarities that can lead to undesirable phenomena such as liquidity hoarding, bank runs and fire sales," they find.
• "Worse, these complementarities will probably strengthen as AI accelerates the adoption of best-of-breed methodologies for measuring and managing risk."

Between the lines: These issues are on the radar of the world's central banks, which view maintaining financial stability as part of their mission.

• A Bank of England staff blog post last month noted that "if multiple firms utilise opaque or black box models in their trading strategies it would be difficult for both firms and supervisors to predict how actions directed by models will affect markets."
• Federal Reserve governor Lisa Cook said in a speech Monday that "the potential widespread adoption of powerful new AI, especially generative-AI applications, inevitably raises questions about potential benefits and risks to the stability of the financial system as a whole."

Go deeper: If you like 22-page papers on systemic risk, the Danielsson-Uthemann paper mentioned above is worthy of your time.