Security Experts Warn AI Deepfakes Threaten Consumers, Voters and Trust in Real Media

Rapid progress in artificial intelligence has made it far easier to produce fake videos and audio recordings, and security experts say criminals and political operators are taking advantage. They warn that hyper-realistic scams and disinformation campaigns could reach people on a large scale.

Deepfakes can copy the voice and appearance of celebrities, political candidates or private individuals, experts told CBS News. According to Doppel, a platform that defends against social engineering and AI-enabled threats, fewer than 10% of people can recognize an audio deepfake. Doppel co-founder Kevin Tian pointed to a test in which only 8% of people realized an AI agent was calling them, while 92% believed they were talking to a human.

The political arena is a particular concern. AI-generated ads featuring deepfakes of politicians have already appeared online and on social media ahead of the midterm elections. In Texas, a manipulated video aimed at Senate candidate James Talarico claimed to show a former student accusing the Democrat of sexual misconduct, according to PolitiFact, a service of The Poynter Institute.

The Wesleyan Media Project, which tracks candidate advertising, reported that as of mid-October campaigns had spent more than $137 million on at least 220 AI-generated ads, some of them containing deepfakes of candidates. Co-director Erika Franklin Fowler said the group found less than $50,000 in such spending during the 2024 cycle, and added that most ad creators do not disclose their use of AI. Fellow co-director Travis Ridout said that even when viewers suspect a fake, the images can still shape opinion. "Even if we know it's not real, that sticks in one's mind," he said.

Raul Liive, a deepfake expert at Veriff, an AI-powered identity verification company, said the technology gets easier to use every day. Years ago, faking a face took effort and image-editing software, he said, but now a person can simply prompt an AI tool. Veriff has recorded a 300% rise in deepfake-related fraud attempts against businesses this year compared with 2025.

Liive described common scams aimed at consumers. In one, a supposed friend calls asking for a loan, even though that person has never made such a request. In another, a caller claims to face an urgent medical bill. He advised people to consider how likely it is that the person they know would really ask for such a thing, and to hang up and call back on a known number or verify identity through another channel.

Spotting fakes is getting harder. In a Veriff test, 70% of people wrongly judged a deepfake video of a female character to be real. Older warning signs, such as extra fingers or hands, are vanishing, and Veriff found that people often rely on unreliable cues like odd skin, hair or teeth, flaws that AI is increasingly able to remove.

Experts also say that growing doubt can damage trust in genuine material. Michael Kaiser, CEO of Defending Digital Campaigns, a nonpartisan group that offers free cybersecurity services to political campaigns, cited an authentic photo of a senator posing with his wife from a hospital bed. It was released amid speculation that the senator had died after a medical emergency earlier this year, yet some internet users suggested it had been manipulated.

"People don't know if what they are looking at is legitimate, or if requests for donations are real," Kaiser said. "They're questioning everything that's going on and that disadvantages and erodes trust in the whole system."