In a latest interview, U.S. Securities and Exchange Commission (SEC) Chairman Gary Gensler once again warned about the risks artificial intelligence (AI) brings to financial markets. Although different versions of AI technology have been used in the industry for many years, and various fraud prevention and compliance measures have been deployed, artificial intelligence can still pose risks to investors' decision-making.
"At a micro level, I think you're absolutely right, using AI to underwrite loans, insurance, etc. can lead to bias," Gensler said.
He explained that a robo-advisor or broker-dealer, for example, must be carefully programmed to optimize the best interests of the client, not the advisor.
At the same time, Gensler said there are also systemic risks, and there could be risks for both financial firms and investors using the same artificial intelligence tools. This can create a "herding" effect in the market, which occurs when groups or institutions all do the same thing or rely on the same tools, rather than acting independently.
"The natural economy will lead to a monoculture, there will be a basic data set or a basic model, and large parts of the financial sector will rely on it ... to do trading and underwriting on that basis," Gensler said. "FICO scoring has been monocultural for decades, so this is nothing unusual or novel."
The FICO credit score is a method of quantifying and assessing an individual's creditworthiness. In the United States, FICO scores are used in 90% of mortgage loan application decisions. Scores range from 300 to 850, with scores in the 670 to 739 range considered a "good" credit score.
Gensler warned that for those who rely on biased and flawed AI models, "if you make a large portion of the market dependent on one data set, then you could have financial instability...Herding could send us off a cliff inadvertently."
In fact, earlier this year, Gensler warned that companies would rely on a few dominant AI models, which would ultimately increase the likelihood of a crisis.