SEC Chair Gensler sounds alarm on risks of large AI-fueled financial models

Artificial intelligence will ultimately be a “net positive” for “efficiency and access in the financial markets,” Securities and Exchange Commission Chairman Gary Gensler said Wednesday. But the financial sector needs to keep in mind that the technology “comes with risks.”

During a webinar with the nonprofit consumer advocacy group Public Citizen, Gensler spoke of the “American affinity for centralization” in everything from cloud providers to search engines. The SEC chair predicted that the same will be true for large AI base models, upon which aggregators with an “insatiable desire for data” will rely.

“The whole financial sector, indirectly, will be relying on those central nodes,” Gensler said. “And if those nodes have it wrong, the monoculture goes one way, well, then there’s a risk in society and the financial sector at large.”

The SEC last July proposed rules to prohibit investment firms from using predictive data analytics, including AI, that put their interests above those of their clients. Those rules followed March 2022 recommendations from the agency’s Investor Advisory Committee, which called for ethical guidelines regarding AI models used by investment firms and financial institutions.

The SEC’s AI rulemaking received swift industry pushback, but Gensler has held steady in his beliefs about the dangers of AI-washing in the financial sector.

Despite the agency’s regulatory push, Gensler noted Wednesday that financial regulators wouldn’t actually have oversight powers for the large AI models from which financial institutions would draw.

With a government that’s used to “regulations and laws around entities and activities,” Gensler said the approach to AI in financial models should be viewed domestically and abroad as a “horizontal challenge” — though horizontal reviews likely still won’t be enough. 

“I think we’ve really got to think [about], ‘How do you keep some diversity in the system?’ And this is diversity of models and diversity of data sources,” Gensler said. “Otherwise, you end up with a pretty fragile system.”

With his hands tied on the more “macro” regulatory pursuits on AI and financial markets, Gensler said the SEC will do its best to address the “explainability” challenge posed by the technology.

“I think the macro issue and financial stability set of issues, it’s really trying to continue to raise the awareness amongst international colleagues about these challenges,” Gensler said. “I noticed that the Chief Justice of the Supreme Court, in his annual report, even spoke about AI. And so I think the awareness is raising, but I’ll stick to our lane of financial services, and particularly securities law.”