Watch the video above as SEC Chair Gary Gensler sounds the alarm on the role of artificial intelligence in finance. In an exclusive interview with Yahoo Finance’s Jennifer Schonberger, Gensler says that “AI is already very much embedded in our capital markets” and warns that it could play a role in future financial crises.
Gensler also discussed a ruling that deemed Ripple Labs’ XRP token can, in some cases, be a security. Gensler told Yahoo Finance that he was “pleased” with the portion of the ruling that determined Ripple could not sell the token to institutional investors, but “disappointed” with the decision that Ripple could sell the asset on public exchanges.
Video Transcript
JENNIFER SCHONBERGER: Chair Gensler, thank you so much for sitting down with me. It’s great to see you.
GARY GENSLER: Great to see you again, Jennifer.
JENNIFER SCHONBERGER: So you sent a warning shot today on AI. Do you think this could be the next financial crisis absent proper protections in place?
GARY GENSLER: I think that it’s quite possible that a financial crisis of the late 2020s or the 2030s, the after-action report will say, oh, we didn’t know that everybody was relying on the same base or foundation model for their mortgage information or their stock market information. And these base or foundation models are called that for a reason. And you’re going to have a lot of applications built on top of them.
There’s not your economic reasons why that will happen. The models are hard to explain. They’re quite complex. And I think we really do need a serious conversation amongst and between financial regulators around the globe because it’s not about any one model or any one market, but it’s really the reliance on these models in our capital markets.
JENNIFER SCHONBERGER: How immediate is this threat?
GARY GENSLER: Well, AI is already very much embedded in our capital markets. Large-asset managers are using it possibly even to listen to this interview here and take the sentiment. And does the stock market go up or down based upon how you ask your questions or your smile? But possibly, the words I use and choose.
And it’s being built into robo advisors and elsewhere. But if you can take all the inputs of the market pricing and tie that into the sentiment analysis following chief executive officers of companies, you can start to see patterns and trade more effectively. But it’s also possible that we end up with dependencies on a base model and end up with, in essence, a monoculture that the markets all go in one way based upon that base model.
JENNIFER SCHONBERGER: You mentioned financial markets are already using this technology. I want to ask you specifically about generative AI, the rise of ChatGPT, Google’s Bard, and layering that into financial services. We–
GARY GENSLER: Baidu’s Ernie in China.
JENNIFER SCHONBERGER: Right, right. You just gave a speech on that. These machines have shown a proclivity to start learning things on their own that they haven’t been taught. And companies don’t really know how they’re doing that. And they’ve also shown the propensity to fabricate information and pass it off as truth.
GARY GENSLER: Hallucinate.
JENNIFER SCHONBERGER: Correct, to use a terminology of the industry. So what is the risk that these machines start embedding biases, fabricated information, into making recommendations and ultimately decisions? What’s the risk that people’s personal financial information all of a sudden gets disclosed publicly? How do we guard against this, Chair Gensler?
GARY GENSLER: So I think these are important considerations that the models might give recommendations or just narrowly influence our behavior as individuals, so-called narrowcasting to us. Advice or communications or products or pricing that extracts more profit for them out of our pocket. So that’s one issue.
You raise the question about bias. If the underlying data that’s being collected reflects the biases already in our society, then the actual algorithms will reflect that bias as well. And then you mentioned deception.
All of these, the SEC is looking at. On the last, fraud is fraud. If a bad actor uses artificial intelligence to try to deceive the public, we’re authorized but also mandated by Congress to go after that, but it’s not without its challenges.
These are sometimes hard-to-explain models. If there’s biases, other parts of the financial sector, like, the Consumer Financial Protection Bureau looks at that. We have a role in that as well. So I think these are rapidly changing times. And those are very good questions.
JENNIFER SCHONBERGER: So you mentioned the challenges. And I know you have asked SEC staff to provide recommendations about the challenges of regulating artificial intelligence. What has staff reported back to you? What have you found? Where are the biggest gaps?
GARY GENSLER: So let me let me cover three. The staff is working on recommendations with regard to conflicts that might arise when an investment advisor or broker dealer is using artificial intelligence or other predictive data analytics– we’re technology neutral. But if they’re using an algorithm to communicate with us, to engage with us, and if they’re considering their own profits or revenues, their own interests, in essence, are they putting the investor’s interests ahead of the advisor’s interests or the other way around?
And we’re going to take a look at the recommendations from staff on that. Secondly, I mentioned fraud and deception. That’s just active. And of course, if the listening public thinks they’ve been defrauded, there’s Tips, Complaints, and Referrals at the SEC.
And then thirdly, as I mentioned earlier, with regard to financial stability, I think that’s really more a conversation amongst the financial regulators. That’s an important conversation as to how do we guard against the crisis of 2027 or 2032 that might arise out of so much of finance relying on one or two base models that, you know, takes us off in the wrong direction?
JENNIFER SCHONBERGER: Do you think that you’re going to need to write specialized rules to govern AI, in addition to applying securities laws already in the books?
GARY GENSLER: At least at the Securities and Exchange Commission, our approach is looking at the outcomes of the use of technology. We’re technology neutral. We at the SEC don’t regulate calculus.
Now, maybe some of your listeners would wish this. You know, we don’t. And we at the SEC wouldn’t propose that we regulate any individual technology, but we look at the outcomes of its use.
But in terms of our mission, protecting investors and facilitating those who want to raise capital and the markets in the middle, the increased use of AI is on our mind and ensuring that we achieve the rules in place and maybe augment them with additional rules, like we’re considering in this conflict area.
JENNIFER SCHONBERGER: So you said augment, so maybe a little bit of extra building on in terms of rules to make sure that investors are protected when it comes to AI?
GARY GENSLER: Absolutely. But it’s really technology neutral. It’s just that using predictive data analytics raises new challenges for which time-tested policy about ensuring that an investment advisor puts your interests ahead of theirs. Well, now, you have an algorithm that might actually, by the math, possibly be considering the investment advisor’s profits or revenues or other interests, and then to guard against those conflicts.
JENNIFER SCHONBERGER: Switching gears, I want to ask you about crypto.
GARY GENSLER: There you go, Jennifer.
JENNIFER SCHONBERGER: You didn’t think I was going to let you off the hook, Chair Gensler, did you? [CHUCKLES] All right, so last week, we got a major ruling from the Southern District of New York on Ripple. How does this ruling change your approach, if at all, your abilities when it comes to regulating crypto and specifically crypto exchanges?
GARY GENSLER: Well, our approach has been the same. It’s about protecting investors, facilitating those who want to raise money, and the markets in the middle. And that’s true whether it’s in other parts of the securities markets or the parts of the crypto markets that are securities. And in terms of the decision, we’re still taking a look at it.
We’re pleased with how the courts addressed itself too that a token for institutional investors is a security, how they addressed themselves to fair notice, disappointed other aspect about retail investors. But we’re still taking a look at that and considering it.
JENNIFER SCHONBERGER: Coinbase is already relisting Ripple. A lot of these crypto exchanges feel like this was a real win for them that the tokens that they list on their exchanges are not securities, based on this ruling. So what do you say to them? You know, are you concerned about potential precedent that this could set, especially for other charges that the SEC has brought against other crypto exchanges?
GARY GENSLER: Look, again, Jennifer, you’re asking good questions. But there’s ongoing litigation with regard to a number of these crypto platforms. And these crypto platforms are co-mingling a number of services that you’ve seen on these crypto platforms where they could well be trading against their customers, bundling up a bunch of services that we would not allow in any other parts of our capital markets.
So the investing public comes first– that’s how our securities laws were first written– companies raising money, and the markets as well and really protecting them and protecting the integrity of the markets. But I’m not able to go into any ongoing enforcement matters.
JENNIFER SCHONBERGER: So it sounds like it’s still sort of an open question. I wonder– there’s been some confusion that’s been generated from this– does it make you reconsider writing customized rules for crypto?
GARY GENSLER: Again, it’s too early. This was, like, three business days ago. But I would note, the agency has rules on the books right now. It’s about full, fair, and truthful disclosure if you’re raising money from the public, and the public is anticipating profit based on the efforts of others.
That’s why Roosevelt called it the Truth in Securities Act. There’s rules on the books about what it means to be a securities exchange, what it means to be a broker, an investment advisor. There are rules on the books.
We’ve also put forward or even adopted rules with regard to broker dealers in this space called special-purpose broker dealers. We’ve put forward proposals with regard to the safeguarding of those assets or even exchanges in what’s called the decentralized finance space. So we’ve done some of that, but again, we’ll continue to consider.
JENNIFER SCHONBERGER: Republicans on the House Financial Services Committee have put forth a discussion draft on a framework to regulate crypto. And within that proposal, they’re looking at laying out guidelines that would direct which firms would need to register with the SEC. They would have you write customized rules.
And I think the biggest point is that they would have a decentralization test to decipher whether a token is a security, a commodity, or both. I’m curious what you think about these principles, especially the decentralization test.
GARY GENSLER: Jennifer, we or asked by members of Congress, both the House and the Senate on a regular basis to look at their various pieces of legislation. And we generally give them our feedback directly rather than through the interviews on the–
JENNIFER SCHONBERGER: So decentralization, good idea, bad idea in general?
GARY GENSLER: Yeah, I’m going to, again, save our comments on any draft legislation for members that are asking us directly. But this field, the crypto field, Satoshi Nakamoto writes an eight-page paper, and at the core of that paper was this concept of moving value on the internet without a central intermediary, a central bank or a commercial bank, but moving value on the internet, whomever she was, whomever Satoshi was.
This field is actually quite centralized. Around each of the tokens, there tends to be a group of entrepreneurs. And you can find a website, and you can look and there’s maybe a Twitter account and somebody that might even have a name of a CEO or a CTO, chief technology officer. Around the platforms, they’re quite centralized as well.
So this is a field that is not immune to the economics of finance. Since antiquity, finance tends towards some centralization. And that’s what we’ve seen happen in the crypto field as well.
JENNIFER SCHONBERGER: The first half of this year was marked by more than a dozen enforcement actions by the SEC when it comes to the crypto space. Should we expect more of the same in the second half of the year? What can we expect from the SEC when it comes to regulating crypto in the second half of the year?
GARY GENSLER: I think over the last six years, we at the SEC have brought, settled, or litigated somewhere close to 150 actions in the crypto field. So I don’t know what the next six months will hold, and I’m not at liberty to speak about any one enforcement action. But we’re really focused on protecting the integrity of the markets, the investing public, capital formation, as I said earlier, and trying to bring greater compliance within this field.
JENNIFER SCHONBERGER: Before we wrap up, Senator Elizabeth Warren sent you a letter today.
GARY GENSLER: She did?
JENNIFER SCHONBERGER: She did. She says, oh, this is news to you.
GARY GENSLER: As I was preparing for giving a talk at the National Press Club and meeting with you.
JENNIFER SCHONBERGER: Yes, yes. And she says she would like you to investigate Tesla for, quote, “possible conflicts of interest, misappropriation of corporate assets, and other negative impacts on Tesla shareholders related to Elon Musk’s takeover.” Are you prepared to entertain this?
GARY GENSLER: Jennifer, I thank you for bringing this to my attention, but again, we’re not able to speak to any potential investigation or an ongoing investigation, whether we have one or not, regardless of who it might be asked about.
JENNIFER SCHONBERGER: All right. Well, Chair Gensler, we’ll have to leave it there. Thank you so much as always for your insight. It was great to see you. Look forward to seeing you soon.
GARY GENSLER: Thank you, Jennifer.
