Long before neural networks entered the trading floor, economists worried about the systemic risks introduced by automation. From the portfolio insurance strategies that contributed to the 1987 Black Monday crash to the high-frequency trading algorithms implicated in the 2010 Flash Crash, the history of finance is peppered with cautionary tales about what happens when machine logic operates faster than human oversight.
Now, a new analysis from the Centre for Economic Policy Research (CEPR) is reigniting that conversation for the generative AI era, arguing that the underlying architectural choices baked into modern AI systems — not just their outputs — carry profound consequences for global financial stability.
The core concern is one that historians of technology will recognize immediately: when competing institutions adopt similar algorithmic frameworks, their decision-making tends to converge. This herding behavior, amplified at machine speed, can transform isolated market shocks into synchronized, system-wide crises. The CEPR discussion situates this risk within the broader question of how regulators should think about AI design — not merely AI deployment.
This marks a meaningful evolution in the policy conversation. For years, financial watchdogs focused primarily on algorithmic transparency and explainability. The new framing suggests that structural choices — how a model is trained, what objectives it optimizes for, how it handles uncertainty — may matter just as much as what the algorithm ultimately decides.
Historically, financial regulation has always lagged behind financial innovation. The Basel Accords took shape decades after the post-war banking expansion; derivatives regulation arrived well after the instruments themselves proliferated. The question now is whether policymakers can get ahead of AI-driven systemic risk before the next major disruption forces their hand.
As AI becomes further embedded in credit scoring, asset management, fraud detection, and macroeconomic forecasting, the architectural foundations of these systems will quietly shape the resilience — or fragility — of the global economy. That is a lesson history suggests we cannot afford to learn too late.