CHARLOTTE — Experts are warning the public to be vigilant as AI-generated deepfakes become more prevalent in social media videos and telephone calls.
Identifying these digital manipulations requires looking for specific errors in spelling, facial textures, and vocal patterns.
Brandi Fox, founder of the AI Business and Career Lab, and Det. Sgt. Patrick Dixson of the Iredell County Sheriff’s Office are working to educate the public on identifying digital forgeries.
The rise of artificial intelligence has led to sophisticated scams, including fraudulent calls to family members that use cloned voices to simulate emergencies.
[ ALSO READ: Should you trust AI with your health questions? ]
Visual inconsistencies are often the first sign that a video has been manipulated. Fox, founder of the AI Business and Career Lab, teaches technology literacy and safe AI usage to adults in underserved communities.
While reviewing an AI-generated video, she noted that the technology often fails to capture background details accurately.
“So the first thing that sticks out to me is the police tape around So, as you see a lot of times with AI, they cannot get spelling right,” Fox said.
Beyond text, observers should examine minute physical details such as tooth shape, skin texture and how a person’s face moves.
Fox noted that AI still struggles to replicate realistic facial movements and the nuances of human emotion in audio recordings.
“They have not learned to a certain extent how to put an emotion,” Fox said. “Their voice is going to sound the same when they’re upset, when they’re happy, when they’re mad, when they’re sad.”
Law enforcement officials are already seeing these tools used in criminal activity. Dixson works with the Special Victims Unit at the Iredell County Sheriff’s Office, where he investigates fraud cases involving artificial intelligence.
“Sadly, we are getting some cases where someone’s grandma or someone else’s family member may call someone and say, Hey, I’m in jail,” Dixson said regarding the use of voice cloning in scams.
Dixson warned that the technology is rapidly evolving, making detection more difficult over time.
“In five or 10 years, it might be darn near indistinguishable from what is real, and that’s why it’s important for us as law enforcement to get data and videos and photos directly from the source,” Dixson said.
To combat phone-based scams, Dixson recommends that families establish a secret code word to verify identities during real emergencies.
He noted that common red flags include callers who demand secrecy, request money via wire transfers or gift cards, or create a false sense of urgency.
“They’ll try and scare you and keep you on the phone and rush you. If the person is unwilling to hang up and they want to keep you on the phone, that’s a huge red flag,” Dixson said.
Fox suggests that users practice critical thinking when viewing any content online, especially when quickly moving through social media feeds.
“A lot of times when we’re doomscrolling, we’re not thinking about, hey, let me stop and look at this video,” Fox said.
She recommends cross-referencing news articles to see if they are posted on multiple reputable pages.
“Use common sense. If it feels fake, if it feels funny, verify before you move forward. I’m really big on what I share,” Fox added.
Major technology companies have implemented measures to address manipulated media.
Meta, which owns Facebook and Instagram, uses “AI info” labels on content detected as being generated by artificial intelligence.
A spokesperson for Google stated the company has clear policies to protect users and govern AI use, noting that its tools are continuously improved to reflect those policies.