The FBI has been warning for more than a year that criminals are using generative artificial intelligence to make their scams more convincing, and one of the most alarming forms is the cloned voice. With only a short sample of someone speaking, often lifted from social media, fraudsters can synthesize a voice close enough to a real family member’s to fool a frightened relative on the phone. The bureau’s recommended countermeasure is disarmingly low-tech: a private family code word that no machine can guess.
The scheme is a modern twist on the long-running grandparent scam, in which a caller claims a loved one is in trouble and needs money immediately. What AI adds is credibility, replacing a stranger’s voice with what sounds like the very person supposedly in danger, and stripping away the accent or hesitation that once gave impostors away.
How a cloned voice powers a distress call
The FBI laid out the risk directly in a public advisory on AI-enabled fraud. In its alert on criminals using generative AI to facilitate financial fraud, the bureau explained that scammers can generate short audio clips that mimic the voice of a victim’s loved one in a crisis, then use those clips to demand urgent payment. The same advisory noted that the technology reduces the time and effort criminals must spend to deceive a target and can smooth over the human errors that once served as warning signs.
A typical call follows a tight emotional script. The synthesized voice claims to have been in a car crash, arrested, or kidnapped, insists the situation is secret, and pushes for money by wire, gift card, or cash courier before the victim can pause to verify. The urgency is deliberate, engineered to keep the target reacting rather than thinking, and the familiar-sounding voice makes the panic feel justified.
The losses behind the warning
The financial toll has drawn the FBI’s attention because it falls heavily on older Americans. According to figures compiled in the bureau’s annual internet-crime reporting, people age 60 and older accounted for billions of dollars in scam losses, with AI-enabled fraud emerging as one of the fastest-growing categories. Reporting on those figures has tied cloned-voice and other generative-AI schemes to a sharp year-over-year rise in the amounts stolen from seniors.
Those numbers almost certainly understate the problem, because embarrassment and confusion keep many victims from ever filing a report. The FBI has repeatedly urged people to come forward through its Internet Crime Complaint Center regardless, since the case data helps investigators map the networks running these operations and gauge how quickly the AI-driven variants are spreading.
Why a code word beats a cloned voice
The defense the FBI keeps returning to exploits the one thing a voice clone cannot fake: shared private knowledge. The bureau’s advisory recommends creating a secret word or phrase with family members that can be used to verify a caller’s identity, a step echoed by consumer-protection reporting. As a report on the family safe-word strategy put it, an AI can copy how a relative sounds but knows nothing about the family itself, so a prearranged password instantly exposes an impostor.
The mechanics are simple to put in place. A household agrees on a word or short phrase that is not posted anywhere online and not easily guessed, then treats any emergency money request that fails the check as fraudulent. Because the safe word lives only in the family’s memory rather than in any voice sample, no amount of cloning gives a scammer access to it.
Verifying a frightening call before sending money
Beyond the code word, the FBI and security specialists point to a set of habits that break the scammer’s momentum. The strongest is to hang up and call the person back on their known number, or to reach another family member who can confirm the supposed emergency, since a genuine relative in trouble can be verified through an independent channel. A roundup of the FBI’s smartphone-scam guidance similarly stresses slowing down and resisting demands for immediate, secret payment, which are the hallmarks of the con.
The payment method itself is another red flag. Requests for wire transfers, gift-card codes, cryptocurrency, or cash handed to a courier are favored by scammers precisely because those transfers are fast and nearly impossible to reverse, and no legitimate emergency requires paying that way. Limiting how much voice and video a person posts publicly also shrinks the raw material available for cloning in the first place.
The through-line in the FBI’s guidance is that the most effective protection is a plan made in advance, not a judgment call made mid-panic. A family that has already chosen a safe word and agreed to verify any urgent money request through a second channel has, in effect, inoculated itself against a cloned voice, no matter how real it sounds on the other end of the line.
This article was researched and drafted with the assistance of AI and reviewed before publication.
More from Morning Overview
- A Colorado wildfire has exploded past 87,000 acres with no containment
- A wildfire near Ouray swelled past 18,000 acres, putting Colorado towns on alert
- The Pentagon’s newest UFO files describe a fish-scaled, potato-shaped object on camera
- Scientists spotted a rare tusked whale alive at sea for the first time, then fired a crossbow at it.
