Serena Sullivan vividly remembers a recent email she received from someone claiming to be the chair of her department at National University in San Diego: He was in a meeting and urgently needed Sullivan’s phone number.
Sullivan was incredulous. Shouldn’t her boss already have that?
“There was a sense of urgency and an appeal to authority,” she told Straight Arrow. Two common scam tactics.
These spammy, scammy messages from someone posing as a friend, relative or coworker requesting personal information or payment on an exceedingly short timeframe are becoming pervasive. And officials worry they will become more common and insidious as criminals turn to generative artificial intelligence.
Scams predate the internet, but cybersecurity experts say the criminals behind them now use advanced AI to perpetrate fraud on a vast scale. This means your inbox may soon be overwhelmed by even more messages from scammers trying to bilk passwords or bank account numbers.
Scammers have always researched their targets. A simple internet search reveals the department in which Sullivan works and the chair. But generative AI helps criminals conduct queries in a fraction of the time and fire off hundreds of messages in an instant.
READ MORE: Congress wants to regulate crypto. But police say the plan fails
The problem is now so acute that the FBI included AI stats in its annual internet crime report for the first time this April. The bureau received more than 22,000 complaints about AI-assisted fraud, totalling more than $893 million in losses in 2025.
Americans reported losing roughly $16 billion to scams and fraud that year, according to the Federal Trade Commission — a 25% increase from the previous year and the highest figure ever recorded by the FTC. Experts say numbers from both the FBI and FTC are undercounts because many victims don’t report their losses.
Why are fake emails suddenly so much harder to spot?
Mike Centrella, former assistant director of the office of field operations for the U.S. Secret Service, said the rise is almost certainly due to artificial intelligence.
“We’re seeing fraud enabled through AI at a large scale at very quick speeds,” he told Straight Arrow.
Centrella’s mother-in-law recently called him because someone claiming to be an IRS investigator said she was about to be arrested.
“I asked her, ‘Did you pay your taxes?’” said Centrella, who is now the head of public policy for the cybersecurity firm Security Scorecard. “When she said ‘Yes,’ I said, ‘There you go.’”
One of the more surreptitious aspects of AI-assisted fraud is that you can’t always tell when AI is used, according to the FBI.
“AI-enabled synthetic content is becoming increasingly difficult to detect and easier to make,” the bureau’s internet crime report says.
Sullivan is a professor of cybersecurity and knew what red flags to look for. The supposed message from her boss, for example, came from Yahoo, not a university account.
But she worries about people who don’t know the signs, especially as fraud proliferates.
How AI employs new skills for age-old scams
Criminals haven’t changed their strategy, cybersecurity experts stress, but AI can quickly churn out messages free of the grammar and spelling errors that once served as obvious tells for spam.
Artificial intelligence can tailor messages to potential victims using information scrubbed from the internet such as someone’s writing or biography, said Jon Stockton, fraud director at Columbia Bank.
AI “can replicate the grammar and tone the victim typically speaks in,” Stockton told Straight Arrow.
Artificial intelligence makes website creation so easy that fraudsters are running out of free domain names, Gilit Saporta, vice president of product for the fraud lab at the software firm DoubleVerify, told Straight Arrow.
“We’re seeing more websites being set up and it’s easier to build a shell company that will look convincing,” she said.
That means fraudsters can create a link to a website that appears authentic and insert it into an email with a few keystrokes, Saporta said.
In fact, cybersecurity experts interviewed by Straight Arrow shared universal advice: Never click on emailed links.
An economy of scale
A September email to Sara Prior claimed a company called Celebrity Management Service could arrange a meeting with rapper Eminem in his home city of Detroit.
In email exchanges over several days, Prior, who lives in northern California, agreed to pay them in small increments.
“I woke up one day and realized it was a scam,” she told Straight Arrow.
By then she already lost $1,200.
While it isn’t clear whether the fraudsters used AI in this case, cybersecurity experts say the technology can make these messages more believable.
“AI can run at multiple targets, providing personalized messages,” said Stephan Tallent, chief sales officer at cybersecurity firm ArmorPoint.
And criminals using artificial intelligence can increase their intended victims by orders of magnitude, said Kathy Stokes, director of fraud prevention programs for AARP.
“They’re using bots instead of humans to do the outreach,” Stokes told Straight Arrow.
How do scammers use artificial intelligence to target businesses?
Scammers also target businesses. If a company lists its clients and employees online, advanced artificial intelligence can, in the blink of an eye, use that information to craft a message.
A Columbia Bank survey in June found that roughly 70% of small and midsized businesses suffered fraud-related financial losses in the past year.
“In past years it was 35% or 40%,” Kathryn Albright, Head of Global Payments & Deposits at Columbia Bank, told Straight Arrow.
She attributed the rise in part to AI-assisted fraud. And while business scams are less likely to work, they are generally more lucrative.
“If they’re trying to compromise the CFO, the payout is going to be a lot bigger if they pull that off,” said Jean Leroux, senior analyst at the social network monitoring service Graphika.
Hong Kong police in 2024 said scammers employed AI to do just that, using a sophisticated deepfake video to mimic a company CFO that tricked a finance worker into sending scammers $25 million.
Police did not reveal the company’s name, but said that several arrests were made in the scheme.
The same year, the cybersecurity firm Knowbe4 acknowledged that North Korean hackers used AI bots to pass employment screenings for remote jobs.
Fighting AI with AI
Law enforcement and large companies also use AI to combat AI-assisted fraud, but it can feel like playing catch-up, cybersecurity experts said.
“The bad guys might have the advantage on law enforcement,” Centrella said. “But that’s going to change over time.”
Investigating scams “is a labor- intensive process,” Daniel Wagner, director of product and engineering for the cloud computing firm Q2 Holdings, told Straight Arrow. In the past, when fraud was relatively rare, it was possible for a human to take on that work. Now it goes to AI.
“It was OK for [a financial institution] to take two weeks to figure out what happened,” he said. “Now we don’t have that luxury.”
How can you protect your personal information from AI scams?
Some advice to consumers and businesses hasn’t changed, Stockton told Straight Arrow.
“We’re seeing a tremendous amount of bank impersonation and law enforcement impersonation,” he said. “The solution for all of those is to hang up and call and verify.”
Customer service can confirm any problems with your bank account, Stockton said.
Virtually everyone has a digital footprint, but proactive measures will reduce the amount of information scammers can find, Sullivan said.
“It surprised me how many people don’t know about privacy settings,” she said. “Just about any popular social media platform, Facebook, Instagram, lets you set your account to private.”
Sullivan also counseled against listing contact information online.
“If you just don’t put that information out there, there’s nothing for them to scrub,” Sullivan said.
Round out your reading
- What does low water at the Hoover Dam mean for power in the American West?
- Could GLP-1 drugs become the first longevity medications? Scientists think it’s possible.
- ChatGPT told him he wouldn’t face charges. Then came the arrest warrant.
- How one small city built a second act as a bio hub inside empty cigarette factories.
- If permanent Daylight Saving Time is so popular, why is it dead in the Senate?