AI can give an old-fashioned impersonation scam a powerful new ingredient: a convincing copy of someone’s voice, face or identity.
Scammers are using the technology to make fraud more believable and easier to produce on a large scale, lawmakers and experts warn.
One of the best defenses is surprisingly low-tech: verify urgent requests through a separate channel and agree on a family safe word.
For years, families have been victimized by scammers pretending to be a loved one in trouble and who needs money. Artificial intelligence is making that scam considerably more dangerous.
The scammer may now sound exactly like your daughter.
That was the frightening reality described by a California woman during a Senate Special Committee on Aging hearing examining how artificial intelligence is changing fraud, particularly fraud targeting older Americans.
Deborah Del Mastro, 64, told lawmakers that a caller claiming to be connected to a Mexican drug cartel said he had kidnapped her daughter. Then she heard what appeared to be her daughter crying and asking for help.
Del Mastro believed it.
She spent about five and a half hours traveling between retailers in the San Francisco Bay Area as she attempted to send money to Mexico. Her daughter, however, hadn't been kidnapped. She was safe.
The voice Del Mastro heard was an AI-generated imitation.
The episode illustrates why lawmakers, banks and consumer advocates are increasingly worried about AI fraud. Scammers no longer have to depend entirely on their ability to tell a convincing story. Technology can help supply the evidence that makes the story seem real.
The scam hasn't changed. The evidence has.
Grandparent scams, government impostor schemes, fake investments and bogus medical products all existed long before generative AI.
What is changing is a scammer's ability to manufacture credibility.
A criminal can potentially pair a familiar fraud script with a cloned voice. An investment scam can feature an apparently recognizable expert. A questionable product can be promoted by a seemingly real doctor. Phishing messages can be customized and polished in seconds.
That distinction was summed up by Paul Benda, executive vice president for risk, fraud and cybersecurity at the American Bankers Association.
“Generative AI is not replacing traditional scams. It is industrializing them,” Benda told the committee in prepared testimony.
In other words, consumers aren't necessarily facing entirely new categories of fraud. They're facing more convincing versions of scams they already know.
And they may be facing many more of them.
A report released by the committee in conjunction with the hearing said Americans reported nearly $21 billion in cybercrime losses during 2025. More than 22,000 complaints involving AI-enabled scams accounted for $893 million in reported losses.
The report also estimated that more than 82% of phishing emails are now created with AI assistance.
Cybersecurity and emerging-technology attorney Matthew Ferraro pointed lawmakers to an industry estimate projecting that generative AI could contribute to U.S. fraud losses reaching $40 billion by 2027.
You can't necessarily believe your eyes, either
Voice cloning is only one piece of the problem.
Dr. David Amron, founder and medical director of the Roxbury Institute, described discovering advertisements that appeared to show him promoting a “miracle” treatment for lipedema.
He hadn't endorsed it.
According to Amron, scammers took material from his YouTube videos and used AI to reproduce his likeness and voice. Even after he reported the ads to platforms, search engines and the site's domain registrar, additional versions appeared.
For consumers, that changes another familiar rule of thumb.
Seeing a recognizable doctor or other expert apparently speaking on camera is no longer enough to establish that an endorsement is legitimate. The same applies to videos involving celebrities, executives, government officials and potentially people you know personally.
That means consumers increasingly need to verify the source of information rather than relying on how authentic it looks or sounds.
Scammers still need you to act
AI may improve the disguise, but it doesn't eliminate one vulnerability in the scammer's plan: The victim still has to do something.
Usually that means sending money, providing information, clicking a link or granting access to an account.
That's why the behavioral warning signs of fraud remain important even when the voice or video seems genuine.
An unexpected demand for immediate payment should raise suspicion. So should instructions to keep an emergency secret, remain on the phone, avoid contacting family members or move money using gift cards, cryptocurrency, wire transfers or other difficult-to-reverse methods.
Bank employees can also serve as a line of defense.
Benda told lawmakers that technology helps financial institutions identify suspicious transactions, but an employee's relationship with a customer can be just as important. A customer who suddenly appears frightened, evasive or coached while making an unusual withdrawal or transfer may be showing signs of exploitation.
AI, in other words, is being deployed on both sides. Criminals can use it to make fraud more persuasive, while banks can use it to detect activity that doesn't fit a customer's normal behavior.
Washington is looking for more than consumer education
The Senate hearing produced bipartisan agreement that telling people to “be careful” isn't an adequate response to the problem.
Lawmakers and witnesses discussed stronger identity-verification systems, faster sharing of information about fraud, cooperation between federal and state authorities, enforcement against organized criminal operations and greater responsibility for platforms where deceptive AI content appears.
Sen. Mark Kelly, D-Ariz., highlighted legislation aimed specifically at AI and older Americans. His bipartisan Aging with AI Act would support research into how the technology affects older adults, while his Senior Chatbot Protection Act would establish disclosures and safeguards involving chatbots and voice assistants.
AI also has legitimate applications that could benefit older consumers, including detecting financial exploitation and supporting health and social services.
The challenge is getting those benefits without giving criminals a nearly unlimited supply of fake identities.
A family code word suddenly makes a lot of sense
Consumers can't be expected to become deepfake investigators every time the phone rings.
Fortunately, they don't have to.
The most useful response to a suspicious emergency call is to verify the underlying claim instead of trying to determine whether a voice is fake.
If someone who sounds like your daughter says she's in trouble, hang up and call your daughter directly. If a caller says he's from your bank, use the number on your card or the bank's official website rather than a number supplied by the caller. If someone claims to represent a government agency, contact that agency independently.
Families can also choose a private word or phrase that everyone knows. If a relative supposedly calls with an emergency, asking for that word provides another layer of verification.
Most importantly, don't allow urgency to become proof.
A convincing voice, familiar face or emotional plea can make an AI-assisted scam feel real. But the pressure to act immediately remains one of the scammer's most useful weapons — and taking the time to verify the story remains one of the consumer's best defenses.