When DeAndrea Salvador began receiving emails in June from people in Brazil, the North Carolina state senator assumed they were part of a phishing scam. The emails, first to her senate office and then to her campaign, had subject lines like “Is this you?” and “AI manipulation.”
But after an email from a journalist asking her about a specific video circulating online, she decided to investigate.
Salvador searched for her name on the internet and found that a clip from a 2018 TED Talk that she had given about ways to reduce energy costs for low-income Americans had been used in an award-winning video to promote Whirlpool products in Brazil.
In the TED Talk, Salvador says that when she was growing up, some of her neighbors resorted to illegally tapping into electricity lines to stay warm in the winter and cool in the summer. She then presented slides with data on how millions of American families, particularly in Black and rural communities, have to choose between paying their electricity bills and other essentials like medication and food.
But in a 20-second clip that appeared in the Whirlpool video, Salvador appears to be talking about low-income residents of São Paulo, Brazil’s most populous city and a world away from where she works on the issue in her native North Carolina. In addition to changing what she says, the spot alters her slides to display data about São Paulo.
In both videos, Salvador appears on the TED stage wearing a sea-green blazer, gray slacks and kitten heels. But her voice has been changed.
At the TED Talk, she said: “People are faced with impossible choices. In the U.S., the average American spends 3 percent of their income on energy. In contrast, low-income and rural populations can spend 20, even 30 percent of their income on energy.”
In the Whirlpool spot, she seems to say: “People are faced with impossible choices. In low-income communities in São Paulo, the average electricity bill costs represents 30 percent of their monthly income.”
The video then jumps to a line she said later in her talk: “This is when energy becomes a burden.”
“I was really shocked,” said Salvador, 34, who founded a nonprofit aimed at reducing energy costs for low-income people before becoming in 2021 the youngest woman to ever serve in the North Carolina legislature.
“They used me as an authoritative figure and turned me into a talking head for Whirlpool, and I don’t even know if what they have me saying is even true,” she said. “And what’s distressing to me is, if no one had come forward, I would have never known that this was out there.”
Salvador has filed a lawsuit in a North Carolina federal court against Whirlpool, Omnicom Group — the global marketing company that created the video — and their subsidiaries over the unauthorized use of her image. The video, which can still be found online, may damage her reputation as a lawmaker and as an advocate for energy affordability, she argues in the lawsuit.
AI manipulation is probably far more common than many people believe and can be hard to detect, said Andrea Hickerson, dean of the journalism school at the University of Mississippi.
“Taylor Swift can call a press conference or put out an Instagram post and say, ‘That’s not me,’” said Hickerson, who studies how AI-generated content is changing the media landscape. But few people know what economic and corporate leaders look like, making them good targets for AI manipulation, she said.
By the time Salvador saw the video, titled “Efficient Way to Pay,” it had already won two awards at the Cannes Lions International Festival in France. Sometimes called the Oscars of advertising, the awards are some of the most prestigious prizes in the marketing industry.
The video created by the São Paulo-based subsidiary of Omnicom Group, DM9, for Whirlpool’s Brazilian branch, Consul, explains how low-income Brazilians could finance the purchase of energy-efficient Whirlpool appliances by using money saved on future utility bills. At Cannes Lions, the judges awarded DM9 and Consul two honors for the two-minute video: their highest award, the Grand Prix, in the creative data category, as well as a Bronze Lion in the creative commerce section.
After Cannes, the video began attracting attention — and questions — about its content. CNN Brazil filed a complaint against Whirlpool and the advertising firm after discovering the video also included manipulated content from a CNN Brazil broadcast.
A spokesperson for Whirlpool said company officials weren’t “aware that Senator Salvador’s remarks had been altered” and contacted her “shortly after we were made aware of the issue at Cannes.”
Omnicom didn’t respond to requests for comment, but on LinkedIn, it’s subsidiary acknowledged using AI to alter the Whirlpool video and others. DM9 apologizes “unreservedly to the Cannes Lions Festival of Creativity, to the judges who took the time and care to evaluate their work, and to the industry as a whole,” the statement said.
DM9 returned Cannes Lions awards it won for three entries, including the Whirlpool spot, due to “serious inconsistencies related to the veracity or legitimacy of the works presented.”
Omnicom and Whirlpool have called for her lawsuit to be dismissed, arguing that Salvador’s reputation hasn’t been materially affected by the video. In a response to Salvador’s lawsuit, DM9 said the video was never broadcast in Brazil and the AI alterations were executed by a small number of employees.
Derek Leben, a professor of business ethics at Carnegie Mellon University’s Tepper School of Business, said the video may be an example of shadow AI, in which employees and outside firms use artificial intelligence without the approval of their managers.
Large companies like Whirlpool should verify the origin of videos, images and data provided by advertising agencies and other companies they work with, Leben said. “I’m skeptical that there’s going to be a law anytime soon that requires disclosure of when generative AI is used to create media but I’m hopeful that companies will demand some kind of transparency from the people they work with,” he said.
But the technology is widely available, making it difficult to protect yourself from deepfakes, said Hany Farid, a computer science professor at the University of California at Berkeley. “Anybody who has any footage of themselves online — a single image, a short video, an audio clip — can have their likeness ripped out, and then people can create audios, videos, and images of them saying and doing things they never did,” Farid said.
Even before she became aware of the video, Salvador said she had already become concerned about the privacy issues associated with the burgeoning world of artificial intelligence. She’s not against its use, she said, noting that she just finished a MBA and fellowship with the Massachusetts Institute of Technology, where she worked on technology that uses AI to address mobility issues in seniors.
“I genuinely don’t think it’s zero sum,” said Salvador, who serves on the National Council on Electricity Policy, a group of state officials who work on energy policies. “We’re seeing rapid innovation that could improve quality of life, but I think with any technology there has to be safe guards.”
She has spent weeks scouring the internet for the manipulated videos of her TED Talk and trying to get those taken down.
“I now have moments of hesitation before agreeing to a speaking engagement if there’s going to be video,” she said.
The post Deepfake of North Carolina lawmaker used in award-winning Whirlpool video appeared first on Washington Post.
