AI voice cloning scams are real enough that American regulators and the FBI have issued formal warnings, yet none of the official sources checked gives a loss total for cloned voices alone. On February 8, 2024, the Federal Communications Commission (FCC) ruled that calls made with AI-generated voices are "artificial" under the Telephone Consumer Protection Act, and the FBI has warned that criminals clone a relative's voice to demand urgent money. A 2025 cross-border grandparent fraud indictment, however, is described in the charging announcement without any mention of AI.
Key takeaways
- The FCC declaratory ruling of February 8, 2024 took effect immediately and applies the same consent rules to AI-generated voices as to other artificial or prerecorded voices.
- The FBI describes criminals using short audio clips of a loved one's voice to pose as a relative in a crisis, and advises agreeing on a secret word or phrase.
- A March 2025 U.S. indictment of 25 Canadian nationals over a scheme that took more than USD 21 million describes human callers, a fake bail bondsman, and no AI.
- A 2023 study found listeners correctly spotted speech deepfakes 73 percent of the time, so judging a voice by ear is unreliable.
- No official figure for losses linked specifically to cloned voices turned up in the sources checked for this article.
What the FCC ruling changed
The FCC announced a unanimous Declaratory Ruling on February 8, 2024. It recognizes that calls made with AI-generated voices are "artificial" under the Telephone Consumer Protection Act (TCPA), which restricts telemarketing calls and the use of artificial or prerecorded voice messages. The agency said the ruling took effect immediately and gives state attorneys general another basis to act, because the use of an AI-generated voice in the robocall is itself covered, not only the scam that follows. A coalition of 26 state attorneys general had written to the FCC in support, and the FCC reported agreements with 48 attorneys general to work on robocall cases. The FCC announcement frames the ruling around calls to consumers; it is not a general ban on voice-cloning software.
What fraud agencies describe
The FBI's Internet Crime Complaint Center published an alert on December 3, 2024 titled "Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud." It says criminals generate short audio clips of a loved one's voice to impersonate a close relative in a crisis and ask for money or a ransom, and that audio of a person may be used to try to reach bank accounts. Its advice includes agreeing on a secret word or phrase with family, limiting public sharing of voice and images, and hanging up to contact an organization through a number looked up independently. A second alert, dated May 15, 2025, reported AI-generated voice messages impersonating senior U.S. officials and said voice cloning "can sound nearly identical" to the real person.
Canada's national fraud reporting service, known as the CAFC, describes the emergency or grandparent fraud without referring to AI: a caller poses as a relative or a police officer, invents an arrest or accident, and presses for cash or cryptocurrency. Its guidance is to hang up and call the relative on a known number, and it notes that caller ID can be spoofed and that the Canadian justice system does not allow bail to be paid in cash or cryptocurrency.
A case where the AI claim does not appear in the record
On March 5, 2025, U.S. Immigration and Customs Enforcement announced that a federal grand jury in Vermont had charged 25 Canadian nationals. The indictment was returned on February 20 and unsealed on March 4, 2025, and it alleges a grandparent scam run from call centers around Montreal between summer 2021 and June 4, 2024, with losses of more than USD 21 million from elderly victims in over 40 states. According to the announcement, callers claimed to be a grandchild arrested after a car crash who needed bail money, some posed as attorneys, victims were told a gag order prevented them from telling anyone, and cash was collected from victims' homes by a person posing as a bail bondsman.
The announcement does not mention artificial intelligence or voice cloning. One public AI incident database entry lists the case with synthetic audio as the alleged technology, but the excerpts from the news reports cited on that entry describe callers claiming to be grandchildren and do not say a voice was cloned. The point is narrow: an emergency scam works without AI, and an AI label added to a case can outrun the documents.
How well people spot a cloned voice
A 2023 study in PLOS ONE by Kimberly Mai and colleagues tested 529 participants in English and Mandarin. Listeners correctly identified speech deepfakes 73 percent of the time. The authors list limits: the clips came from an older text-to-speech model, all used one female speaker, participants knew the purpose of the task, and clues such as knowing the speaker were removed. Results with newer systems and real calls may differ. The finding supports the FBI's call-back and secret-word advice more than advice to listen for flaws.
Claims against records
| Common assumption | What the checked record shows |
|---|---|
| AI voice calls to consumers are unregulated | The FCC treats AI-generated voices as artificial voices under the TCPA, effective February 8, 2024 |
| Voice cloning is behind the biggest grandparent frauds | The 2025 Vermont indictment announcement describes scripted human callers and does not mention AI |
| Fraud losses from cloned voices are measured | The Federal Trade Commission's June 15, 2026 release reports USD 3.5 billion lost to imposter scams in 2025, with no AI breakdown |
| A careful listener can hear the fake | Listeners in the 2023 study spotted deepfakes 73 percent of the time |
The FTC release also reports total fraud losses of about USD 16 billion in 2025 and says imposter scams made up nearly one in three fraud reports. Those totals include all impersonation methods, from fake bank alerts to calls posing as the government.
Steps that fit both the FBI and Canadian guidance
- Hang up and call the person back on a number already saved, not on one given during the call.
- Agree on a family code word in person, and do not post it online.
- Treat any request for gift cards, cryptocurrency, cash pickup, or secrecy as a warning sign.
- Do not trust caller ID, which can be spoofed.
- Report the call through the CAFC reporting page in Canada or to the FBI's Internet Crime Complaint Center in the United States.
This article gives general information and is not legal advice. In Canada, section 403 of the Criminal Code makes it an offence to fraudulently personate another person to gain an advantage or obtain property, with a maximum of 10 years' imprisonment when prosecuted by indictment. Questions about a specific incident belong with police or a lawyer.
FAQ
Is cloning someone's voice illegal?
The rules covered here attach to how a voice is used, not to the software. Robocalls with AI voices fall under the TCPA in the United States, and fraudulent personation is an offence in Canada under section 403 of the Criminal Code. Using a voice without permission can raise other claims that depend on the jurisdiction.
Can the FCC ruling stop a scammer in another country?
Not by itself. The ruling gives U.S. regulators and state attorneys general a legal basis against AI-voiced robocalls, but enforcement against callers abroad depends on carriers, cooperation, and investigators.
Do these scams always use AI?
No. The 2025 Vermont case relied on a scripted story, a second person posing as a lawyer, and in-person cash collection. Cloned audio makes the story harder to doubt, but the pressure, secrecy, and payment demands are the same.
What is the single most useful precaution?
Verifying through a second channel. Both the FBI and the CAFC advise hanging up and reaching the person on a number already known, and the FBI adds a pre-agreed secret word.
Related reading
- Wider harms attributed to AI
- The threat or tool debate
- Ethical questions raised by AI
- How Canada regulates AI
More in AI Policy and Ethics.
Last verified: October 2026
Sources
- FCC Makes AI-Generated Voices in Robocalls Illegal, Federal Communications Commission, February 8, 2024
- Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud, Alert I-120324-PSA, FBI Internet Crime Complaint Center, December 3, 2024
- Senior US Officials Impersonated in Malicious Messaging Campaign, Alert I-051525-PSA, FBI Internet Crime Complaint Center, May 15, 2025
- Emergency scam page, CAFC (Canada's national fraud reporting service)
- 25 Canadian nationals connected to nationwide multi-million dollar grandparent scam charged in Vermont, U.S. Immigration and Customs Enforcement, March 5, 2025
- FTC Data Show People Reported Losing $3.5 Billion to Imposter Scams in 2025, Federal Trade Commission, June 15, 2026
- Mai KT, Bray S, Davies T, Griffin LD, Warning: Humans cannot reliably detect speech deepfakes, PLOS ONE, August 2, 2023
- Criminal Code, section 403, Justice Laws Website, Government of Canada