With the advancement of AI technology, sophisticated scams using AI voice technology are on the rise. Engineer Tim Green has compiled a summary of these cutting-edge AI voice scams.
The Three-Second Theft: Why AI Voice Fraud Outruns Every Defense — SmarterArticles
https://smarterarticles.co.uk/the-three-second-theft-why-ai-voice-fraud-outruns-every-defence
Sharon Brightwell, a resident of Hillsboro County, Florida, USA, was a victim of a sophisticated fraud scheme using AI voice cloning technology in July 2025.
Brightwell described the scam call he received as ‘like hearing my daughter crying.’ The call began with an AI voice clone of Brightwell’s daughter claiming to have hit a pregnant woman while driving and that his cell phone had been confiscated by the police. The call then switched to a man claiming to be his daughter’s lawyer, who explained that $15,000 (approximately 2.5 million yen) in cash was needed as bail. The lawyer warned him ‘not to mention the purpose’ when withdrawing the money from the bank, explaining that it ‘could damage his daughter’s reputation.’
Within an hour of receiving the call, Brightwell withdrew cash and handed it over to a delivery person who appeared to be affiliated with the court. He later contacted his real daughter and realized that she had not been in the accident, meaning the call was a scam.
The AI voice fraud that Mr. Brightwell fell victim to is now not only one of the most common crimes in America, but also one of the most technologically sophisticated.
In April 2026,
the FBI ‘s Internet Crime Complaint Center released its ‘2025 Annual Report on Online Crime,’ which for the first time in the 26-year history of the report included ‘AI-driven fraud’ as a separate category. According to the report, more than 22,000 complaints related to AI were recorded in 2025, with losses exceeding $893 million (approximately 146 billion yen) due to AI crime. Of this, victims aged 60 and over accounted for a staggering $352 million (approximately 58 billion yen) in losses, revealing that the elderly are the most targeted age group in AI-driven financial crimes.
The FBI pointed out that the figures in the report only include ‘those that victims recognized and reported,’ and that it underestimates the true extent of the problem. Green explained that since most victims of AI voice clone phone calls are unaware that AI is involved, it is more appropriate to interpret the FBI’s reported ‘$893 million in losses due to AI crime’ as a lower limit, not an upper limit.
The core technology behind AI voice scams, which are spreading primarily among the elderly, is the ability to generate synthesized voices indistinguishable from the original using just three seconds of audio data. Because only ‘three seconds’ of audio data is needed, it can be generated from voicemail messages, podcast excerpts, or videos posted on Instagram accounts. Therefore, Green explains, ‘A grandchild appearing in just one TikTok video provides all the information a scammer needs to fabricate a case.’
Furthermore, one reason why the threat of AI voice fraud is becoming more serious is not only because AI voice clones can be easily created, but also because the tools for doing so are very inexpensive and abundant. In addition, Consumer Reports has pointed out that most voice cloning products such as Descript, ElevenLabs, Lovo, PlayHT, Resemble AI, and Speechify lack effective security measures against misuse and abuse, highlighting the lack of security measures in these technologies.
ElevenLabs, a particularly prominent AI voice cloning tool, claims to have built a multi-layered security program. In fact, ElevenLabs includes a usage policy prohibiting impersonation, a public AI voice classifier that can identify voices that may have originated from its own system, a tracking function that links generated content to the account that created it, and a ‘banned voice’ protection function that prevents the creation of cloned voices of specific protected individuals during election periods.
Regarding this, Green pointed out that even ElevenLabs’ AI voice cloning technology, which boasts industry-leading security systems, falls short in terms of protection. ‘ElevenLabs’ security systems are not trivial and are superior to those offered by many competitors. However, these measures have a structural weakness: almost all of them are reactive. They help investigators identify the source after the fraud has occurred and the victim has lost their savings. In other words, they do little to prevent the generation of an AI voice clone from a three-second audio clip. This is because what is needed to prevent cloning—a rigorous and enforceable mechanism to detect cloning—is something that a highly competitive and rapidly changing market is reluctant to impose on itself. If protective measures cause companies to lose profits, the market will not voluntarily offer them,’ he said.
The reason older people are targeted in AI voice scams is that they have higher average savings and are an efficient target. Once a call is successful, they can earn far more profit efficiently than if they targeted younger people. ‘Even if you explain 100 times that ‘voices can be faked,’ that knowledge goes out the window when an AI voice clone pretends to be your child or grandchild and asks for help,’ Green said.
Furthermore, Green points out that one way to regulate AI voice fraud is to require users of the tool to provide proof of identity upon starting to use it, in order to detect ‘who created the voice clone using AI.’
Click Here For The Original Source.
