Election year starting to show cracks from “nuclear” disruption of deepfakes
The packed election year of 2024 is ramping up and so are political attacks using deepfakes. U.S. President Joe Biden has already been faked in robocalls related to primary elections in New Hampshire. Now deepfakes have surfaced in more local races, with victimized politicians comparing the potential impact of deepfake technology on democracy to weapons of mass destruction.
For firms working at the edge like Paravision and ID R&D, the emergent threat has spurred action to develop new tools that can match fraud and injection attacks powered by AI, video deepfakes, audio spoofs and other weaponized technology. Biometric authentication, facial recognition, algorithmic deep learning and liveness are among the countermeasures available. But the most precious resource is time: news of new attacks breaks weekly, and the biggest fish – the U.S. presidential election – is months away.
In the U.S. alarm bells are ringing at both state and federal levels of government, prompting a scramble to enact regulatory legislation. Spectrum News New York reports on a fake audio recording circulating in which former state Assemblyman Keith Wright was heard disparaging sitting Assemblywoman Inez Dickens by calling her lazy and incompetent. Wright, who sent a cease-and-desist letter to the platform hosting the fake content, compares the current regulatory landscape for deepfake spoofs to the Wild West. “It’s just something that’s very, very dangerous,” he says. “It’s almost like a nuclear bomb that people are able to use.”
Audio “hot mic” deepfakes among the most compelling and dangerous
A report from CNN has disturbing echoes of the New York story. This time the setting is last year’s Chicago mayoral race. Deepfake audio of candidate Paul Vallas made the rounds, in which Vallas espoused extreme pro-police opinions. (“Back in my day, cops would kill, say, seventeen or eighteen civilians in their career. And nobody would bat an eye.”) Vallas lost the election by four points. “I won’t be the first and I won’t be the last” to be the victim of a deepfake scam, he tells CNN.
Hany Farid, a digital forensics expert at UC Berkeley, says these kinds of “hot mic” audio deepfakes could be even more compelling than faked video. “You don’t see the mouth moving,” says Farid. “You don’t see anything. But you hear the voice, and it’s visceral. It sounds like you’re eavesdropping on them. I think those are really powerful.”
ID R&D recently released new software for detecting voice cloning and preventing deepfake audio attacks, in response to developments like these.
Laws and tools fly in effort to dismantle algorithmic time bomb
Regulators have also been working to keep up. PBS reports that this week, the Federal Communications Commission (FCC) outlawed robocalls that contain AI-generated voices. Those robocalls from deepfake Joe Biden in New Hampshire are now illegal under a technicality of the 1991 Telephone Consumer Protection Act.
“Bad actors are using AI-generated voices in unsolicited robocalls to extort vulnerable family members, imitate celebrities, and misinform voters,” says FCC chairwoman, Jessica Rosenworcel. “We’re putting the fraudsters behind these robocalls on notice.” Violating the law could lead to lawsuits or fines of up to $23,000 per call.
Authorities in New Hampshire have identified the source of the calls as Texas-based Life Corp. and its owner, Walter Monk, with the calls being transmitted by Lingo Telecom. Both entities have previously been investigated for separate instances of illegal robocalls. Meanwhile, states including California and Wyoming are among those pursuing or already enforcing laws prohibiting the distribution of synthetic media without disclosure and with the intent to manipulate or deceive. In New York, following the Keith Wright deepfake, Queens Assemblyman Clyde Vanel is pushing for new laws to fight what he calls “identity theft on steroids.”
Voting rights groups are similarly concerned and pushing for even speedier legislation on deepfakes. Organizations such as Public Citizen and the Brennan Center for Justice have lobbied Congress for a crackdown on synthetic media and biometric deepfakes and tougher penalties for anyone using the technology to try and fool voters.
Having been transformed by social media, electoral processes now stand to be invaded en masse by synthetic humans with bad intentions. As the results play out, very little will be certain – except that it stands to be a wild and disorienting trip.