YouTuber’s voice was cloned by AI without consent to promote videos

Imagine you are at a garden party, say 30 or more, with your significant other. If you lost sight of that person, could you recognize them by sound alone?

Probably yes for most of us. That’s because our voice is an important part of who we are and is unique to each of us. That uniqueness was largely true until the advent of generative AI. YouTube content creator Jeff Geerling learned this the hard way.

YouTube influencer’s voice cloned

Geerling is a programmer and tech educator who has been a full-time YouTube content creator since 2021, but began posting high-quality, edited videos in 2019. Over the past five years, he has built a significant following of over 769,000 subscribers. In an email exchange, he wrote that through his videos, he aims to educate interested viewers about tiny Raspberry Pi computers, software programming, and electronics in general.

On Sunday, he posted a video on his channel titled “They Stole My AI Voice.” In it, he explained that an electronics manufacturer called Elecrow had used AI cloning software to copy his voice, then repeatedly used AI scripts to copy Geerling’s voice in about two dozen product tutorials. One of his subscribers pointed out the uncanny similarity between Geerling’s unique voice quality and the tutorials’ voiceovers.

“At first I was in shock; I thought it was me, but I went to my wife and kids and asked if they would like to listen to the recording and see what they thought of it. Everyone, including my three-year-old, thought it was me and didn’t understand why I was even asking,” Geerling wrote.

“I also felt a bit betrayed because I had reached out to Elecrow multiple times in the past and they even offered me an ongoing business partnership since my YouTube audience is in the same industry they target with their products,” he added.

The company sends the YouTuber an apology and offers to pay

Geerling then filed a complaint via the company’s website about the situation, and on Monday morning he received an apology email from Elecrow CEO Richard Lee. Geerling posted a redacted excerpt of the email on his social media platforms and agreed to share it below.

Geerling wrote that he was impressed by Lee’s quick reaction and acceptance of responsibility.

“But I think Richard and the marketing staff were sincere in their apology and handled the situation very well. Many in a similar situation would try to deny wrongdoing or avoid responsibility, but they properly acknowledged the situation and said they understood the seriousness of it,” Geerling explained.

He added that Elecrow offered to compensate him for using his voice, but that it wasn’t a bid for money. Instead, he decided to go public to draw attention to the potential copyright infringement and identity theft risks that AI technology poses today, with virtually no regulation.

“It’s difficult to protect against AI voice and image theft with the tools we already have today. However, AI video and speech generation services may require more stringent security measures beyond just a terms of service checkbox when creating voice or video clones,” Geerling wrote.

Elecrow did not immediately respond to an emailed inquiry for this article.

AI Detection Tools for Voice and Audio

On Monday, my article on detecting AI-generated video provided a few tools to determine if a clip was produced using AI. Here are two tools for AI-generated voices: the first is the AI ​​voice detector, and the second to consider is the Eleven Labs AI speech classifier.

Geerling’s voice-cloning issue is just the latest in a series of similar cases. In May, actress Scarlett Johansson accused OpenAI of cloning her voice to launch a conversational AI assistant called Sky. OpenAI CEO Sam Altman has denied the claim.

In February, ahead of the New Hampshire primary, a political consultant used AI to create fake, automated phone calls from then-candidate Joe Biden urging residents not to vote in the primary.

Recently, a British bank notified its members that fraudsters only need three seconds of audio gleaned from a social media video to create a believable AI voice. Using the same social media account, bad actors can direct friends and family to send money using the AI-created voice. Despite this ongoing trend, people are largely unaware of the scam.

Last month, 46% of respondents to a survey conducted by the bank were unaware that such AI voice scams could take place, while almost 10% admitted they would send money to help a family member or friend in need — even if the call seemed strange. It should come as no surprise that more than a quarter of those surveyed said they had fallen victim to an AI voice clone in the past year.

A simple way to protect yourself from phone scams using artificial intelligence

The best way to protect yourself against this type of AI voice cloning attack is to have an easy-to-remember, super-secret password that you can get from family and friends who you can send money to in an emergency.

So if you get a call from a crying teenager or your aunt who claims a shady tow truck is holding her car ransom unless you immediately send $200 via Venmo, calming down and asking the person on the other end for their password before transferring the funds could save you a lot of hassle.

Like any tool, AI voice cloning can be used for good or ill. But few things are as naturally human and personalized to each of us as our voices. Now that thieves can use AI to capture our images, written words, and spoken words, you wonder—and shudder—if AI might one day be used to capture our thoughts.

I’d rather not think about it.