A the last robot call Fake President Joe Biden telling New Hampshire residents not to vote was almost artificially generated, according to disinformation experts and people who study the technology.
The origin of the call, which the New Hampshire Attorney General’s Office described Tuesday as an apparent “illegal attempt” to prevent voters from writing in on behalf of Biden in the state’s Democratic primary, is unknown. Experts say it appears to be fake audio or video created by artificial intelligence and designed to mimic real people, usually without their knowledge or consent.
“All the signs point to this being deeply fake,” said Ben Colman, CEO of Reality Defender, which makes software to check whether media files are artificially generated.
“We never say something is 100% certain because we don’t have the ground truth, but it’s highly likely to be manipulated,” Colman said.
voice in robo call, first obtained by NBC News, sounds like Biden’s, though the cadence is off. It’s almost impossible to determine which AI program may have generated the audio; programs that can create a moderately convincing copy of someone’s voice are widely available, such as phone apps and online services, usually for free or for only a small fee.
Such programs often need only a small amount of sample audio to reproduce people’s voices, making it easy for them to impersonate a politician.
Lindsay Gorman, who researches emerging technologies and disinformation at the German Marshall Fund’s Alliance for the Security of Democracy, said that although technology is constantly improving, information is often deeply falsified.
“The cadence seemed unnatural, robotic, especially towards the end. Here’s one clue to a potentially fake piece of audio content,” he said.
Gorman said clearly defining what is or isn’t deeply fake becomes a cat-and-mouse game with developers improving the technology.
“In the case of visual deepfakes, one signature is being able to see people’s eyes,” he said. “This is generally good information, but if their eye movements are unnatural, it can be a deep fake. But now deepfakes has gotten even better and fixed some weird erratic eye movements: staring, blinking a lot.”
Sen. Richard Blumenthal, D-Conn., who introduced the AI legislative framework in the Senate, said he hopes the incident will alert Americans to the dangers of misinformation that AI can pose.
“It can be mastered by a neophyte in minutes. “Unfortunately and tragically, if we don’t act decisively, this is our future,” Blumenthal told NBC News.
“My hope is that Biden’s profound fraud or impersonation will actually be a shock to the system and maybe alert everyone that actually everyone is at risk,” he said.
while federal law criminalizes knowingly attempting to prevent people from voting or registering to vote, there is little regulation of the deceptive use of AI. while some activists While Deepfake has pressed the Federal Election Commission to regulate ads, it has not yet determined whether to begin any rulemaking process around the technology, the spokesperson said.
Mekela Panditharatne, senior counsel at New York University Law School’s Brennan Center for Justice Democracy Program, said the apparent use of deep-fake technology to keep people from voting in US presidential primaries could be as new as the use of robocalls. not to suppress voices.
“Robocalls have historically been used by fraudulent actors to spread false information about how, when and where to vote,” Panditharatne said. “Voice generation AI potentially makes the method of attempting to suppress voices more attractive to fraudsters.”