SEC votes to develop AI, cyberproposal for policy overhaul amid growing concerns
Republican commissioners worry new policies could limit adoption of useful technology
{{#rendered}} {{/rendered}}The Securities and Exchange Commission (SEC) on Wednesday adopted new rules to address cybersecurity and artificial intelligence (AI) concerns.
"I think for the SEC, probably the biggest concern there is potentially, how does this actually impact individual investors and any sort of other aspects of that," Peter Klimek, director of technology at cybersecurity firm Imperva, told Fox News Digital.
Commissioners voted to propose a rule that would require broker-dealers to address conflicts of interest in the use of AI in trading, which already had potentially dangerous consequences when it came to light during the 2021 GameStop "meme stock" event that they may have utilized AI to amplify user behavior.
{{#rendered}} {{/rendered}}The new rule would require companies to disclose a cyberbreach within four days after determining if it resulted in serious material consequences for investors, and to periodically describe their efforts to identify and manage threats in cyberspace.
WHAT IS ARTIFICIAL INTELLIGENCE (AI)?
SEC headquarters in Washington, D.C. (Graeme Sloan / Bloomberg via Getty Images / File)
The AI proposal would require broker-dealers to "eliminate or neutralize" any conflict of interest that occurs if a trading platform’s predictive data analytics put the broker’s financial interest ahead of that of the firm’s clients.
{{#rendered}} {{/rendered}}"We've seen instances where already efforts to use large language models for automating security tasks have resulted in instances where the hallucinatory effects of the models have basically led to false positives," Klimek said.
AI, AUTOMATION FORECAST TO TAKE MORE JOBS FROM WOMEN THAN MEN BY 2030
SEC Chair Gary Gensler speaks during an interview in Washington, D.C., on July 27, 2023. (Andrew Harrer / Bloomberg via Getty Images)
"I think this ability to lead you down the wrong path is really something that's not unique to any individual domain," he said. "That's really where I think we see some of the risks associated with the models, especially as organizations look to really turn these into actual products that they can deliver to users and consumers."
{{#rendered}} {{/rendered}}Klimek highlighted the volume of data required to train AI models as a concern since it likely will lead to even more bots appearing on the internet to scrape data, raising privacy concerns.
ACCELERATED ADOPTION OF AI COULD AUTOMATE 30% OF AMERICANS' WORK HOURS
Traders work on the floor of the New York Stock Exchange on Aug. 26, 2022, in New York City. (Spencer Platt / Getty Images)
Even more concerning, though, would be the automation of tasks that could lead to AI attempting to manipulate stock markets and prices.
{{#rendered}} {{/rendered}}"Talking about the context of stock prices, the manipulation, this is potentially an area where you can see automated bots really trying to influence markets that certainly will have the potential as well," he said, suggesting that analysts need to pay attention to finding vulnerabilities in software and the potential impact on data breaches.
Republican commissioners objected to these first two measures, claiming the proposal was unnecessary in light of brokerages' disclosure requirements and could stifle the use of new technologies.
CLICK HERE TO GET THE FOX NEWS APP
{{#rendered}} {{/rendered}}In a third vote on Wednesday, the SEC unanimously proposed the requirement for more internet-based investment advisers to register with the federal agency, narrowing an exemption that officials said some had used to avoid this.
If adopted, the rule would require that investment advisers provide investment advice through a functioning, interactive website, among other changes, thereby preventing them from using the two-decade-old exemption inappropriately.
Reuters contributed to this report.