Deepfake scams infiltrate social media as voice cloning becomes easier

WASHINGTON, D.C. (InvestigateTV) — Your voice may no longer be your own.

Scammers have found ways to steal the sound, tone, and inflection of everyday Americans’ voices and turn them into AI deepfakes – and most people may not even know it.

A new survey by the cybersecurity company McAfee finds that deepfakes are infiltrating social feeds, most commonly on Facebook, YouTube, TikTok, and Instagram. Experts say scammers are using AI voices against victims for financial fraud and to falsely impersonate them, and it only takes seconds of audio to do it.

Some deepfakes are playful — a seal jumping on a trampoline or a dog driving a car. Others are more serious. Politicians and campaigns are posting deepfake ads, like one a rival posted of Georgia Sen. Jon Ossoff using his face and voice to make controversial statements and sharing he’s only seen a farm on Instagram.

“A deepfake is a piece of media, audio, video, images where a real person is either AI-generated or AI-manipulated using tools that are available to anybody with just a Google search,” said, Ben Colman, CEO of Reality Defender, a deepfake and AI-generated media detection company.

But deepfake audio and video clips are no longer limited to famous people. Experts and victims say they are seeing an increase of everyday people being targeted as the technology becomes more accessible.

Everyday people targeted

Brewster County Sheriff Ronnie Dodson never expected he would be deepfaked. The Texas lawman learned someone took old YouTube videos of him and made a viral AI deepfake video using his real voice to endorse a health supplement.

Licensed cosmetologist Karen Flowers in Virginia had a similar experience. With over 90,000 followers on her YouTube channel, someone took her image wearing a custom shirt from the hair care tutorials she posts online and had her selling life insurance.

“I knew it was a deepfake,” Flowers said.

YouTube eventually agreed and shut down the fake channel after months of reaching out and reporting the video to the platform. But the cases involving her and Dodson underscore how realistic AI-generated content can appear — and how difficult it can be to catch.

Technology advances rapidly

Colman said making deepfakes has become dramatically easier.

“I’d say four years ago you had to be an AI expert. You had to be a computer science background. You had to know how to code. And now my 8-year-old son can make a deepfake,” Colman said.

Colman founded Reality Defender five years ago. The company helps banks like JPMorgan Chase & Co, insurance and health organizations, and government organizations detect those impersonations before fraudulent activity occurs.

The McAfee survey of thousands of Americans found that one in 10 have now experienced a voice clone scam - and about 53 percent say they share their voice online at least once a week.

“If I wanted to make a deepfake of you, I would simply go on Google, look up your name, I’d find a photo of you, perhaps on social media or LinkedIn, and then I would find audio of you, typically only needing three to five seconds,” Colman said.

He said what makes voice cloning particularly dangerous is that the majority of transactions and approvals still happen through voice.

“And so in a world where you’ve been told for the last decade that your voice is your password and now anyone can recreate any voice,” Colman said.

Colman said it’s a security risk because it only takes three seconds of audio to match a voice, and there are hundreds of apps already out there that can make audio deepfakes.

Harder to detect

Detecting deepfakes continues to become more challenging, according to Colman.

“In the last year has gotten so good that even the PhDs on my team with their eyes can’t tell the difference,” Colman said.

Alex Nette, CEO of cybersecurity company Hive Systems, said the technology has improved so quickly that older detection methods may no longer work as reliably.

“As AI gets better, that appearance, that replication only gets better and better. So, as we used to see things like face drift, where maybe it was trying to look like a person, but the face was kind of slipping off of that person’s face. That is all going away,” Nette said.

Nette said that just as phishing emails became more polished over time, AI scams are following a similar path. He said more consumers are now being targeted with basic voice clones.

Easy targets

Nette and Colman both stressed that the growing accessibility of the technology is putting more people at risk.

“Maybe a phone call that sounds exactly like me. The same cadence, the same way I talk. My voice could call one of my employees, ask them to buy $500 in gift cards and send all the codes to them, which is a classic scam,” Nette said.

Colman said business owners are particularly vulnerable.

“Many business owners are going to be scammed by the use of this AI, and while they may not be direct targets of AI or scammers, they will end up being caught up in this dragnet,” Colman said.

Deloitte’s Center for Financial Services documented deepfake fraud losses of $12.3 billion in 2023 and predicts that they could climb to $40 billion in the U.S. by 2027.

Little regulation

“We are confident that regulations will catch up, but right now it’s absolutely the wild west,” Colman said.

Colman drafted a bipartisan proposal and testified before the Senate Judiciary Committee in April of 2024 to explain how regulations could be strengthened and how consumers could better identify AI-generated content.

One recent federal law, the Take It Down Act, was passed by Congress and signed into law in 2025. It makes it a crime to share intimate images without a person’s consent, including deepfakes video, but not audio.

Detecting the technology

Colman said his company expects to make the technology available to consumers within the next 12 months to help identify those videos.

“We’re updating our solution typically every week, very much like antivirus software. It is a cat-and-mouse game, but I would say that AI gives more advantages to the defenders than to the attackers,” Colman said.

Experts suggest the following if you think a call, video or social media post may not be real:

• Hang up the phone and call the person back if you suspect that something seems suspicious.
• Create code words with family, or ask them something AI could not guess, such as what did they ate for dinner last night.

“I teach my kids to ask, hey, what we have for dinner last night. Because, you know, AI is not going to know what you had for dinner last night,” Colman said.

• When scrolling social feeds, if something seems unbelievable, ask yourself is this legitimate.
• Do research and don’t rely on watermarks on videos because scammers are now manipulating those, too.