Amazingly, the “CFO” image and voice were computer-generated, as were those of the other executives who appeared on the call. And the accounts belonged to scammers. The worker was the victim of a stunningly elaborate artificial intelligence scam, according to local media reports. The millions remain lost.

Welcome to the dark side of AI technology, in which the voices and faces of people you know can be impeccably faked as part of an effort to steal your money or identity.

Scientists have been programming computers to think and predict for decades, but only in recent years has the technology gotten to the level at which a computer can effectively mimic human voices, movement and writing style and—more challenging— predict what a person might say or do next. The public release in the past two years of tools such as OpenAI’s ChatGPT and DALL-E, Google’s Gemini (formerly Bard), Microsoft’s Copilot and other readily available generative AI programs brought some of these capabilities to the masses.

AI tools can be legitimately useful for many reasons, but they also can be easily weaponized by criminals to create realistic yet bogus voices, websites, videos and other content to perpetrate fraud. Many fear the worst is yet to come. We’re entering an “industrial revolution for fraud criminals,” says Kathy Stokes, AARP’s director of fraud prevention programs. AI “opens endless possibilities and, unfortunately, endless victims and losses.”

Criminals are already taking advantage of some of those “endless possibilities.”

▶︎ Celebrity scams. A “deepfake” (that is, a computer-generated fake version of a person) video circulated showing chef Gordon Ramsay apparently endorsing HexClad cookware. He wasn’t. Later, a similar deepfake featured Taylor Swift touting Le Creuset. The likenesses of Oprah Winfrey, Kelly Clarkson and other celebs have been replicated via AI to sell weight loss supplements.

▶︎ Fake romance. A Chicago man lost almost $60,000 in a cryptocurrency investment pitched to him by a romance scammer who communicated through what authorities believe was a deepfake video.

▶︎ Sextortion. The FBI warns that criminals take photos and videos from children’s and adults’ social media feeds and create explicit deepfakes with their images to extort money or sexual favors.

Eyal Benishti, CEO and founder of the cybersecurity firm Ironscales, says AI can shortcut the process of running virtually any scam. “The superpower of generative AI is that you can actually give it a goal; for example, tell it, ‘Go find me 10 different phishing email ideas on how I can lure person X.’ ”

Governments are scrambling to keep up with the fast-evolving technology. The White House in late 2023 issued an executive order calling for increasing federal oversight of AI systems. It’s a technology, it noted, that “holds extraordinary potential for both promise and peril.” That led to the establishment of the U.S. AI Safety Institute within the U.S. Department of Commerce to “mitigate the risks that come with the development of this generation-defining technology,” as Commerce Secretary Gina Raimondo put it.

As it turns out, AI may be our best tool for countering the malicious use of AI.

Benishti’s company develops AI software that detects and prevents large-scale phishing attempts and ransomware attacks. AI also is a key tool for detecting suspicious transactions at your bank, for example, flagging unusual charges on your credit card and blocking scam calls and texts.

The problem, says Craig Costigan, CEO of Nice Actimize, a software company that develops technology to detect and prevent financial fraud, is that “most of these scams and frauds are done by folks using the exact same tools as we use—but they don’t have to abide by the rules.”

AI technology also is used to tackle robocalls, says Clayton LiaBraaten, Truecaller’s senior strategic adviser. “If we see phone numbers generating hundreds of thousands of calls in a few short minutes, our models identify these patterns as suspicious. That gives us a very early indication that a bad actor is likely behind those calls.”

Banks use predictive AI as well. Costigan’s company, Nice Actimize, creates AI-based software that financial institutions use to sift through vast amounts of data to detect anomalies in individuals’ patterns, he explains. “It could be that someone is withdrawing $50,000, which is an unusual amount. It could be the location of the IP address. Why is the transaction happening in London?”

What’s possibly more alarming is voice cloning in an industry that for so long has used verbal confirmation to authorize transactions, Costigan says. “Fraudsters can call up and say, ‘Hi, move this money for me.’ And that voice sounds exactly like you. That’s a problem today.”

Banks are considering going beyond voice confirmation, so “you may also get a single follow-up question, like what’s your favorite color,” Costigan says. “They may now even require something additional that validates that you are you.”

Consumers have a role in protecting themselves, Benishti says, by understanding that “they cannot 100 percent trust communication, especially unsolicited.” Fraud fighters need to be ready to adjust their strategies as scammers are “very astute technologists and accomplished psychologists,” with evolving techniques, LiaBraaten says. “It’s a cat-and-mouse game,” he adds. “We just have to stay ahead of them.”