Swindlers Are Now Using AI. Here Are the Warning Signs.

Artificial intelligence is making financial scams harder to spot.

Interactive deepfakes such as live phone calls or videoconferences with what seems to be a real person are becoming more common. In February, a Hong Kong company lost HK$200 million (U.S. $25.6 million) after employees fell for a deepfake of the firm’s chief financial officer ordering money transfers during a video call.

Also in February, Gizmodo reported that the Federal Trade Commission has identified thousands of AI scams impersonating Elon Musk and his companies, Tesla, SpaceX, Neuralink and X, as fraudsters attempt to swindle people out of money. The story highlighted one person in Pennington, N.J., who lost $18,880 during the Tesla Cybertruck online event.

Lou Steinberg, managing partner at CTM Insights, a cyber research lab, says the interactive deepfakes are especially problematic since “you tend to trust people when you are talking to them and getting responses.”

A Gallup survey from November showed 15% of respondents said someone in their household fell victim to a financial scam, and 57% worry about being tricked by a scammer into sending money. That is higher than the 51% who are worried about someone breaking into their car.

The figure lines up with Stuart Sprenger, senior wealth advisor, personal wealth management at Citi, experienced. He says over the past few years about 10% of his clients have lost money from scams, and those are the ones he knows about.

How they work

The deepfake calls will impersonate financial institutions, asking a potential victim to verify accounts or read back pin codes so they can steal money, while some may fake family members’ voices to trick someone into sending funds. “Microsoft says it can recreate your voice in three seconds,” Steinberg says.

In November, Philadelphia corporate attorney Gary Schildhorn testified before a Senate special hearing on AI and fraud, describing how he nearly fell victim to scammers using AI to clone his son’s voice. The scammers said his son was in a car accident and needed $9,000 to post bail.

Steinberg says AI can personalize phishing in large numbers of emails and texts, giving scammers a bigger target. The days of obviously fake emails rife with typos are going away, and instead consumers are receiving fake, personalized communications that won’t automatically land in their spam basket. Fake website links are also more sophisticated.

“They know where you work, because you put it on LinkedIn. They know where you vacation, because you posted that on social media. They know the name of your kids,” he says.

It isn’t just scammers reaching out. Joanne Bradford, chief money officer at Domain Money, a financial advisor, says AI makes fake merchants harder to detect. Bradford, who was the former president of online shopping savings platform Honey, says before buying anything, verify the merchant through social media, Google reviews or even calling the business.

“Anything can be replicated very, very quickly with AI right now. Provenance and authenticity from who you’re really buying from is very critical,” she says.

Fighting back

Recouping stolen money is difficult, so the experts urge people to take preventative steps such as reviewing financial accounts often, using two-factor authentication, and limiting the amount of information you give out.

Steinberg says his family members have a code word they will use if they are worried that an urgent call from a family member is fake. He also suggests not to answer “yes” to questions verifying identity since those can be recorded and used elsewhere.

Interactive deepfake calls are scripted, so one way to detect AI-generated calls is to throw the computer off-script. Steinberg says a friend of his challenges what he thinks are AI robocalls with non sequitur responses. Not long ago, Steinberg said a robocall asked his friend to verify personal information to possibly hack into an account. “’So, he said, ‘Last week, you ran over my duck.’ The AI doesn’t know how to respond to that,” he says.

Veronica Perez, manager of loss prevention at Affinity Federal Credit Union, agrees that scammers often work off a script, and they have scraped enough personal information off websites to make the person they are contacting believe the communication is legitimate. By speaking with a sense of urgency and using just a small amount of public information, the AI scams trick people into revealing more information.

“Pay attention to what information they’re providing to you. Are they using your name? If they say there’s unusual activity on your Visa card, ask them, ‘What kind of card?’ Have them provide the last four digits of the card number,” she says, adding legitimate institutions reaching out will have that information.

Don’t be afraid to hang up and call back, she adds, even if the caller is pressuring the victim to stay on the line. “They’re trying to avoid you from thinking, ‘Does this make sense?’” she says. “If it’s secretive, or if you’re being told to lie, or prevent sharing that information with friends or family or law enforcement, that’s a huge red flag.”

Write to [email protected]