How AI Can Cause a Financial Crisis

If it is not enough that AI might destroy the human race and take over jobs, the technology could also wreck the financial markets.

That is according to Gary Gensler, the chair of the U.S. Securities and Exchange Commission. He told the Financial Times that the increasing use of advanced AI systems and concentration of power in a few AI platforms would risk a financial crisis in the coming decade.

Many financial institutions might be “relying on the same underlying base model or underlying data aggregator” that is “sitting not at the broker dealer, but it's sitting at one of the big tech companies," he reasoned. The result could be “herd” mentality if parties base decisions on the same data model, thereby undermining financial stability.

Gensler's viewpoint is gaining traction among industry experts as well.

“There is a reasonable chance that uncontrolled AI could lead to disruptions in financial markets,” Piyush Tripathi, lead engineer at the financial services platform Square, said in an interview. “AI, like any tool, relies heavily on the data it's built on. Unsatisfactory or biased data could result in poor or even dangerous recommendations, potentially leading to a market crash.”

AI Crash?

This perspective is not new for Gensler. In 2020, he co-authored a research paper titled "Deep Learning and Financial Stability," in which he expressed a similar viewpoint. In this paper, he collaborated with Lily Bailey, who was then a research assistant at MIT and now works at the SEC as an assistant to the chief of staff, according to her LinkedIn profile.

They wrote that the growing adoption of artificial intelligence systems in the financial sector "may result in financial system fragility and pose risks to the broader economy."

The paper advocates for government regulation by suggesting that "existing regulatory frameworks in the financial sector, developed in an earlier era of data analytics technology, may be insufficient in addressing the systemic risks arising from widespread adoption of deep learning in finance."

To be sure, AI holds remarkable promise for financial markets as well, said Tripathi. Automatic trading systems, some of which are already using a form of AI, reduce emotional influences in market decisions, potentially leading to more stable market behavior. The AI systems also can work round the clock, constantly analyzing data and generating valuable insights, which can transform market operations.

“Yet, we must be wary of potential negatives,” he added. “Beyond the hallucinations and biases, there is a risk of AI systems becoming so complex that they are difficult to correct if anything goes wrong.”

AI has already crashed markets, or at least flash-crashed them a couple of times, Maya Mikhailov, the founder of SAVVI AI, said in an interview. High-frequency trading combined with algorithmically driven strategies have been identified as the culprit of more than one flash crash over the last couple of years. A flash crash is a rapid decline and recovery in the market.

“Could a misinformation campaign powered by realistic deep-fakes be used to crash individual stocks or sectors?”  Mikhailov added. “It has likely already occurred, just not at the grand scale of the "whole market." If anything, regulators and market makers are now more aware of these risks with AI, putting systems in place with circuit breakers, and civil and criminal penalties to address these risks.”

Promising, yet complex, future

Not everyone believes that AI will lead to financial ruin. Giuseppe Sette, president at Toggle AI, which makes AI-powered investing software, said in an interview that Gensler’s concerns are not wholly founded.

“The industry is running at 100 miles per hour,” he added. “He needs to be the one to say, 'Guys, there is a speed limit. We need to do this properly.' We need someone whose sole utility function is to make sure we don't screw up. He is going to be a good sounding board for us. We welcome a conversation with the regulators.”

Mikhailov asserted that AI will now become the most influential market force in the future, just as digitalization has driven the conversation in the markets for the last 20 to 30 years.

That is because only AI can analyze the vast amount of data generated by companies, industries and markets at the speeds modern markets require.

“Finding signals through that much noise is exclusively the domain of machines. Now, where humans can still outwit markets is forecasting in low or no data scenarios where no previous information exists, or black swan events, where a low-probability scenario comes to pass,” Mikhailov added. “But even then, they will likely use AI-driven analysis to understand the downstream effects of those black-swan occurrences.”

The future of AI in finance is promising yet complex, Tripathi said. As AI systems become more sophisticated and their use in financial decisions expands, we can expect them to play an increasingly prominent role in shaping market behavior.

“But these advancements come with potential risks, and it is crucial to maintain robust oversight and conscientious data practices to prevent issues,” he added.