Lisa Rozner joined CBS News New York in 2017. She covers a wide range of stories.
You may have seen ads recently about "the pink salt diet," the latest weight-loss gimmick with glowing endorsements from celebrities and doctors.
CBS News New York's Lisa Rozner has found that some of those ads are AI generated, known as deepfakes, and bad actors are using the likeness of one Manhattan psychologist, to purport a phony endorsement.
Dr. Rachel Goldman educates the public about obesity regularly and reached a pinnacle in her career when she appeared alongside Oprah Winfrey on "Oprah Daily," starting back in 2023. She has since appeared on several episodes.
But that worldwide exposure has also made her the target of digitally manipulated, AI-generated deepfake videos. She said she's worried some may be taking the harmful medical advice, thinking it's coming from her.
Goldman, who is also an assistant professor at NYU Grossman School of Medicine, said it has been challenging to get the videos taken down, so she's issuing a warning to the public.
In one episode back in 2023, Goldman told Winfrey about speaking to people struggling with obesity.
"We don't want to define an individual by their medical condition," she said.
From that same episode, an AI-generated version of her was somehow created and she says seen by her followers on Facebook and YouTube with her instead saying, "The pink salt trick it's 100% natural and free of side effects."
"It looks like me. It's kind of me talking, but it's not me at the same time," Goldman said. "It's very weird and scary."
But even more scary Goldman says is that people are falling for it. She showed CBS News New York a text from a relative that read, in part, "Does it work? Send me the recipe."
Goldman says at least two people e-mailed her that they purchased the product, including a woman from Australia.
One customer messaged her that she was billed triple, and asked, "Do you have a product or did they AI the whole thing?"
"Every day I'm getting messages," said Goldman. "I want them to know, do not purchase this. It is not something real that I'm endorsing. It is fake.
"I'm really just trying to get back to as many people as I can and spread more knowledge before more people purchase it or try something that could potentially be dangerous, right? We just don't know," she added. "Supplements are not FDA regulated. I tend to say that anything that's a quick fix or a hack usually is not true."
Goldman says the fake AI version even shared false medical credentials, like that she attended Stanford, when in reality she studied at Farleigh Dickinson University.
She said she's worries about the dangers of fake medical endorsements like this.
"Some people are even sharing their medical history with me in an email or in a phone call asking if it's safe for them," Goldman said.
The task of telling the truth and taking deepfakes down is insurmountable, even for a media mogul like Winfrey. A rep said she has a digital security team that works on takedowns of false ads and channels every day.
CBS News New York came across the AI-generated ads while on YouTube. In one, a video from Winfrey is manipulated to make her say, "Just this morning I looked at the exact spot where I keep my pink salt recipe."
Winfrey wrote on Oprah.com in August, "Let me say this clearly: If you see an ad with my face on a 'PRODUCT,' it's fake."
Meta and YouTube told CBS News New York they removed the accounts and ads related to Goldman and Winfrey that violated their policies.
In 2023, CBS Mornings co-host Gayle King took to social media to denounce a deepfake AI video falsely depicting her endorsing a weight-loss product.
CBS News New York asked King, "What was that like for you?"
"Well, think about this, Lisa; I didn’t even know about it. People started calling me asking me where can I get it?" King said. "And then I watched it and I thought ‘Well that is the outfit I was wearing. That is it sounds like my voice. It looks like me.
"But if you look closely but you gotta look really close to see that the lips are slightly off and the way I was speaking isn’t words I would use. But it was very unsettling to me," King added. "I just would hate for people to buy something because they think I’ve endorsed it when I’ve had nothing to do with it and I don’t know is the thing safe you know? What are the ingredients. I knew nothing about it."
Congresswoman Yvette Clarke, D-N.Y., is a senior member of the Energy and Commerce Committee. Since 2019, she has introduced and reintroduced "The Deepfakes Accountability Act."
"We're not protecting the American people the way we ought to," Clarke said. "What my legislation would simply do is require a digital watermark on any deepfake manufactured so that people know and can distinguish immediately that this was generated by artificial intelligence."
Next May, the bipartisan "Take It Down Act" takes effect, and will give websites and online platforms 48 hours to remove fake content.
Attorney Susan Jacobsen specializes in defamation and data privacy cases at the law firm Brennan Manna Diamond.
"This act is going to require covered platforms, which are public websites, and mobile applications like Facebook, Instagram, Twitter, to establish a notice and take-down process," Jacobsen said.
Jacobsen advises victims to screenshot or screen record the videos of any deepfakes and document the harm it causes. She says victims can sue for damages and seek court orders for takedowns if they can identify the perpetrators.
Rozner spoke to three different tech experts who showed her how easy it was to create an AI deepfake video of her.
Matt Swider, editor-in-chief of the tech review website TheShortcut.com, showed Rozner a platform that offers "instant voice cloning," and another that syncs lips.
He took a live report of her from Times Square and entered a new script; the AI-generated version of her was ready in less than an hour.
Rozner noticed her voice was deeper in the manipulated video.
Swider responded, “If I uploaded more video of you, I would be able to get an exact match, but it was only about 20 seconds of video.”
Lionsgate Network helps victims of financial fraud duped by AI. Founder and CEO Bezalel Eithan Raviv says in the past six months, cases have spiked.
"So [bad actors] learn from each and every interaction online," Raviv said. "This is not about hacking a computer or hacking a code. This is about hacking your consciousness."
Kwindla Hultman Kramer, the CEO of "Daily," creates virtual avatars for education tutors and customer service programs.
He agrees that consumers are being targeted in environments they least expect it.
"The very large social media companies where you would happen upon these videos and in a casual watching you might not be looking for the signs of a deepfake," Kramer said.
The FBI says its Internet Crime Complaint Center received more than double the amount of complaints about deepfake AI videos so far this year compared to the same time period last year.
The agency also has statistics showing financial losses have nearly tripled.
"This is an issue that can be dealt with through the FCC or through the Department of Homeland Security but has not been made a priority," Rep. Clarke said.
As for deciphering medical advice online, Goldman says,"I always like to tell people, pause before we react, right? Pause, inquire, ask people, ask your doctor. Like honestly, don't email me. Ask your health care professional, is this something that you should be taking."
As for how to spot a deepfake AI, experts say you should look for awkward facial movements, lip-syncing issues, a person's hand movements and any unnatural speech patterns.
If you're unclear of the validity of a medical endorsement, experts say look up the doctor's official website and social media to see if they've announced the endorsement on their pages.