Back to feed
News Story
The AI Insider
1 sources

Can AI Trading Cause a Market Crash? New Research Shows

New research from Stanford, the Deutsche Bundesbank, the European Central Bank, and the University of Naples Federico II suggests that the design of AI investment systems could influence whether automated investors amplify financial runs. Simulations showed that Q-learning agents sometimes withdrew excessively, while large language models were less prone to runs but made less predictable collective decisions. The study does not prove AI will cause a market crash, but it highlights potential risks as AI is increasingly used in investing.

SynthePulse Insight · AI deep readingMembers

Will AI Trading Trigger a Market Crash? New Research Reveals Mechanisms and Uncertainties

Version 1 · 1 source

A simulation experiment conducted by Stanford University, Deutsche Bundesbank, the European Central Bank, and the University of Naples Federico II suggests that the internal design of AI investment systems may influence whether they amplify financial runs, fail to coordinate, or maintain market stability. However, the research does not prove that AI will cause a market crash, and it has significant limitations.

  • Q-learning agents, after experiencing defaults during training, may redeem excessively even when the fund is in good condition, creating a 'hot stove effect.'
  • LLM agents (DeepSeek-R1) typically do not redeem when the fund is in good condition, but when multiple outcomes are plausible, they fail to coordinate and develop divergent beliefs.
  • Providing LLM agents with slightly different private signals helps them coordinate, while identical information may lead to disagreement.
Open section navigationResearch Background and Core Questions

Research Background and Core Questions

What happens when a large number of autonomous machines make interdependent decisions? This is the question researchers from Stanford University, Deutsche Bundesbank, the European Central Bank, and the University of Naples Federico II sought to answer. Their study focuses on the behavior of AI investment systems during periods of financial stress, exploring whether their internal design influences whether automated investors remain calm, collectively flee, or scatter.

The study used a simplified simulation scenario: a mutual fund holding assets that cannot be quickly liquidated, where each AI investor must choose between redeeming early (receiving a smaller but certain payment) and staying invested (potentially receiving a larger future return). The return depends not only on the fund's fundamentals but also on the behavior of other investors—if enough investors redeem, the fund is forced to sell assets, reducing the value for remaining investors.

Free for now

Read the full analysis

4 more sections of analysis, plus the full takeaway

Loading

Credibility boundary

This article is based on a report by The AI Insider, which cited the researchers' findings. The research itself is a simulation experiment, not real market data. All specific descriptions of AI behavior come from the researchers' reports and are source claims rather than independently verified facts.

Sources for this version

  1. Can AI Trading Cause a Market Crash? What New Research ShowsThe AI Insider

Primary report

The AI Insider

Primary source