Emerging Threats to the U.S. Financial System

3. Deepfakes. AI can now generate audio and video clips that are so realistic many people can’t tell that they’re fake. A few years ago, scammers mimicked the voice of a European CEO—down to his faint German accent—and tricked his colleague into wiring them $243,000. Adversaries could use deepfakes to compromise major companies or business leaders. They could also fan economic crises or controversies with fake outrages.

4. Bond dumping. China holds more than $800 billion in U.S. Treasury bonds. It could, in theory, dump them in an effort to rock U.S. markets and drive up interest rates. That would briefly make it more expensive for the U.S. to borrow money. With perfect timing, that could—again, in theory—trip up the U.S. response to a Chinese provocation, such as a move on Taiwan.

None of these is going to bring the financial system to its knees anytime soon. But that’s not necessarily the point. All of them could make the markets seem a little less safe, a little less secure. They could undermine public trust and reinforce the idea that what we think is real might just be a deepfake or a cleverly orchestrated meme campaign. That would worsen one of our great national maladies, a condition RAND calls “Truth Decay,” the diminishing role of facts and analysis in public life.

“It comes back to this idea of trust in each other, trust in the markets, trust in institutions,” Sytsma said. “That trust is actually critical to how markets function. When you start to lose that, you run the risk of greater market volatility and greater market losses.”

The experts RAND consulted were divided over how much danger these threats actually pose right now. AI manipulation? Possible. Meme campaigns? At least feasible. Bond dumping? “Highly improbable,” the researchers concluded. Russia and China toyed with the idea during the 2008 financial crisis. But China buys U.S. bonds to support its growth and bolster its currency. Dumping U.S. bonds now would be like shooting its own economy in the legs.

But deepfakes? Those already represent a “significant risk” to financial markets, the experts warned. One good deepfake video could briefly disrupt the market, they concluded. But a coordinated campaign of deepfakes could cause deeper and more-lasting harm—measured not just in trust, but in dollars.

Last year, for example, a photograph started spreading on social media that appeared to show an explosion at the Pentagon. Officials debunked it as a fake almost immediately. But the S&P 500 still dropped by .3 percentage points in a matter of minutes, a loss of potentially billions of dollars. And that was just from a photograph.

Advances in AI are making it much cheaper and easier to produce convincing deepfakes. Efforts to counter them are advancing as well; a model developed by Intel looks for biological markers like blood flow to separate real humans in videos from computer-generated fakes. But, for now, the deepfakers have the edge. In a 2020 competition, the winning deepfake-detection model got it right just 65 percent of the time.

Financial regulators have taken steps in recent years to protect the markets against deepfakes and other threats. They have called for greater monitoring of meme stocks, for example, and warned that too many institutions rely on the same AI models, leaving them all open to manipulation. They have sought to prevent sudden market swings by instituting “circuit breakers” that shut down trading when index losses exceed 7 percent.

But more could be done, the researchers wrote. Stronger policies against the misuse of AI in general could provide important safeguards for the market. More research into how Truth Decay could corrode financial systems would also help. But the threats are always evolving and advancing. The researchers recommended periodic exercises—economic wargames—to look for vulnerabilities and test whether existing defenses still hold.

That is a potentially $2.7 trillion concern. That’s what a 7 percent market loss would look like before the circuit breakers can kick in and stop the damage.

Doug Irving is a communications analyst at RAND. This article is published courtesy of RAND.