AI and Economic ‘Doom Loops’: Market Instability Risks Demand Regulation

In December 2023, a report from the Financial Stability Board and the Bank for International Settlements (BIS) highlighted AI's growing role in financial markets. The report flagged a concerning trend: AI-driven trading algorithms create feedback loops that magnify volatility during downturns. These loops threaten entire sectors.

Originally, 'doom loop' described cycles between sovereign debt crises and failing banks. In AI, it refers to machine-learning systems that reinforce trends without sufficient human oversight. AI models trained on historical data amplify herd behavior, converging on similar strategies. In volatile conditions, this convergence can turn routine corrections into cascading sell-offs.

Andrew W. Lo, a finance professor at MIT Sloan, noted, "AI systems don’t just observe markets—they act on them. Their decisions shape the data future models will train on, creating a reflexive loop." Lo spoke during a panel hosted by the International Monetary Fund (IMF) in October 2023.

Evidence of these dynamics is mounting. During the March 2020 market crash, automated trading accounted for over 60% of trading volume on major exchanges, according to the World Federation of Exchanges. Researchers at the European Central Bank (ECB) found that algorithmic trading exacerbated intraday price swings in sovereign bond markets during that period. Their findings, published in Economics Bulletin, identified poor model calibration and lack of cross-market coordination as key failings.

AI's impact extends beyond trading. Machine-learning systems influence sectors like supply chain logistics, energy markets, and climate finance. These systems rely on pattern recognition, which can hard-code biases when conditions shift too rapidly for models to adapt. For instance, AI models used by shipping firms during 2021-2022 worsened global supply chain disruptions by prioritizing short-term cost savings, according to a 2022 study by McKinsey.

The regulatory landscape lags behind technology. Existing frameworks like the EU's Markets in Financial Instruments Directive II (MiFID II) govern human-led trading but lack clarity on autonomous systems. Meanwhile, the US Securities and Exchange Commission (SEC) has yet to finalize its update to Reg SCI, which would extend oversight to include AI-driven systems critical to market infrastructure.

Eswar Prasad, a senior fellow at the Brookings Institution, remarked, "We’re dealing with emergent behavior. No single actor designs these doom loops, but the systemic impact is real. Regulation needs to move from static compliance checks to dynamic supervision models compatible with machine-learning systems."

Some jurisdictions are taking proactive measures. In Singapore, the Monetary Authority of Singapore (MAS) has implemented a regulatory sandbox for AI applications in financial markets, allowing firms to stress-test algorithmic models under controlled conditions. However, such frameworks remain exceptions rather than norms globally.

Efforts to mitigate AI risks increasingly emphasize explainability and transparency. Without a clearer understanding of model decision-making, regulators struggle to intervene effectively. The EU's proposed Artificial Intelligence Act, expected to be finalized in 2024, mandates rigorous documentation and auditing for high-risk AI applications, including finance. Critics, however, point out that the act’s enforcement mechanisms remain untested.

Geopolitical competition complicates matters. While the US and EU focus on accountability, China prioritizes scale. In August 2023, the People's Bank of China (PBoC) unveiled its AI-driven economic forecasting platform, which uses vast datasets to coordinate policy interventions. This system's opacity has drawn criticism from Western analysts concerned about embedded bias.

The stakes are high. AI-driven systems are reshaping markets and influencing government responses to crises. Misaligned models could amplify systemic risks in climate finance or hinder equitable resource allocation during emergencies. As Prasad cautioned, "The irony is that AI, a technology designed to optimize, could become a source of chaos if left unchecked."

The unresolved question is whether regulators can swiftly implement adaptive frameworks. As of late 2023, most major economies have yet to harmonize standards or mandate real-time monitoring tools. The pace of AI adoption accelerates, and without intervention, the economic 'doom loops' described by the FSB may shift from theory to reality.