Forecasts for our future world

Why Smarter AI Could Make Financial Markets Less Stable

Why Smarter AI Could Make Financial Markets Less Stable

A flock of birds can turn almost as one. Each bird responds to nearby movement, yet the group produces a pattern far larger than any individual. Financial markets can show a similar kind of collective behavior: thousands of separate decisions combine into a sudden rush to buy or sell. This is a useful starting point for understanding AI and financial stability. Artificial intelligence may help investors make better decisions. But what happens when many intelligent systems make similar decisions at the same time?

Recent events make the question concrete. In its 6 November 2026 market update, CoinShares reported that the previous day’s sell-off had pushed the Nasdaq down 5.5% and Bitcoin down just over 5%, with more than $1 billion in leveraged crypto positions liquidated, or forcibly closed. The episode illustrates markets moving together; it does not establish that AI trading systems caused the fall.

To understand the connection, think of markets as a network: points joined by links. The points represent investors, banks, and trading platforms. The links represent relationships such as loans and shared investments. Researchers use related mathematical tools to study financial networks, ecosystems, and other systems whose parts depend on one another.

A fund that loses money on technology shares might sell Bitcoin to raise cash. Its selling can push Bitcoin’s price down, affecting investors who never owned those shares. Losses have traveled through a financial connection. Alternatively, both assets might fall because investors react to the same news. Prices moving together do not, by themselves, reveal which mechanism is operating.

Physics offers another useful idea: feedback. Hold a microphone too close to its loudspeaker and a small sound can become a piercing squeal. The microphone picks up the amplified sound, sends it back through the speaker, and strengthens it again. A financial feedback loop can work similarly. An investor buys assets using borrowed money, known as leverage. If prices fall, a lender may demand extra money or collateral. The investor sells assets to meet that demand. Those sales can push prices lower, forcing other investors to sell too. This is positive feedback: “positive” means that the process reinforces itself, not that the outcome is good.

Imagine a deliberately simple model. An initial $100 of forced selling triggers another $80, which triggers $64, and so on. Each round becomes smaller. If the same $100 instead triggers $120, followed by $144, the rounds grow. The mathematical question is whether the feedback weakens or amplifies a disturbance. Real markets have changing prices, buyers, and limits, so these numbers illustrate a mechanism rather than predict a crash.

This mechanism already matters without AI. In its September 2026 record, the Bank of England described how forced reductions in borrowed investment had amplified July’s falls in AI-related and semiconductor shares. It also reported that market functioning remained orderly, without broader systemic stress. A sharp price fall is not automatically financial instability; the deeper concern is disruption to functions such as lending, funding, and payments.

AI could change the strength and speed of these interactions. Trading systems may use similar data, pursue similar goals, or face similar risk limits. They need not share identical software to reach the same conclusion: sell now. Like the flock, coordinated behavior can emerge without a leader.

Mathematicians describe quantities that tend to move together as correlated. Trading decisions can be correlated too. When reactions differ, some participants may buy while others sell. Strongly correlated selling can overwhelm available buyers, producing a larger price move. The danger comes from how decisions interact, even when each decision makes sense for its owner.

On 30 June 2026, Bank of England deputy governor Sarah Breeden warned that AI agents responding similarly to the same triggers could amplify market volatility during stress. She described work on simulations to investigate what drives this herding behaviour.

Yet AI does not inevitably make markets less stable. Federal Reserve researchers found that AI agents in laboratory-style investment experiments often relied more on their own information than on following the crowd. Under different instructions, however, herding emerged. These experiments suggest that behaviour depends on the setting and incentives; they do not settle how an entire market will behave.

The scientific challenge is therefore to test the whole system. How do trading models respond to one another? Do cash reserves absorb losses, or do borrowing rules force everyone to sell? Could temporary trading pauses interrupt dangerous feedback? AI may help explore these questions by simulating many interacting participants. A market’s resilience depends on more than the intelligence of its traders. It also depends on the connections between them, the strength of their feedback loops, and whether their decisions leave room for others to act differently.

Did you enjoy this article?


Cosmologa
Cosmologa

PhD in Theoretical Physics. DeFi, DeSci, DePhy, techno-anarchy, watcher of the Universe. Amateur writer and science communicator. Check out my new science outreach YouTube channel https://www.youtube.com/@CosmologaLabs


Forecasts for our future world
Forecasts for our future world

Trying to predict the future with AI, one question at a time

Publish0x Publish0x

Reward the author with $0.01 in crypto, and earn yourself as you read!

20% to author / 80% to me.
Rewards are FREE. Publish0x pays them, not you.

Page not displaying correctly?