Biden deepfake in NH sounds alarm on AI in elections: New law proposed

HANCOCK — At 5:35 p.m. on Jan. 21, just two days before New Hampshire’s first-in-the-nation primary, Gail Huntley received a phone call with President Joe Biden telling her not to vote.

"I didn't think I was hearing him right," said Huntley, 73, of her immediate reaction. A Democrat from Hancock, she was planning to write in Biden's name on the New Hampshire primary ballot.

“It’s important that you save your vote for the November election,” the voice on the other side of the phone said. “Voting this Tuesday only enables the Republicans in their quest to elect Donald Trump again.”

Huntley knew then that something wasn't right.

“I knew it was a scam," she said. "They had copied his voice."

The voice sounded like Biden and even used his phrase, "what a bunch of malarkey," but it was artificially generated. Huntley hoped that others would also realize that it was fake. “I was just hoping that folks getting that call would know to disregard that message and get out to vote."

This Biden deepfake used to dissuade people from voting in the Democratic primary in New Hampshire underscored the risk AI poses in political advertising and election interference in the US and abroad. In the New Hampshire House this session, lawmakers have introduced House Bill 1596 as one effort to combat the quickly emerging technology.

The Biden deepfake in the New Hampshire primary

The call Huntley received went to thousands of others, according to anti-robocall application Nomorobo.

The scam used a tactic called spoofing, in which scammers make the call look like it’s coming from another number to obscure their identity. In this case, the number was Kathy Sullivan’s, who is a former state Democratic Party chair and helps run the super PAC Granite for America, which supported the write-in Biden campaign. At least a dozen people called Sullivan after receiving the call.

On Tuesday, John Formella, the New Hampshire attorney general, announced his office's Election Law Unit identified the source of these calls as Texas-based Life Corporation and an individual named Walter Monk. He credited the Anti-Robocall Multistate Litigation Task Force, which is a bipartisan group of 50 state attorneys general, and the Federal Communications Commission Enforcement Bureau.

"The partnership and fast action in this matter sends a clear message that law enforcement, regulatory agencies, and industry are staying vigilant and are working closely together to monitor and investigate any signs of AI being used maliciously to threaten our democratic process,” Formella said.

The Election Law Unit sent a cease-and-desist order to Life Corporation for engaging in voter suppression.

“That was a crime,” said New Hampshire Secretary of State David Scanlan. "The robo call with Biden's voice may be the first instance where there was an artificial intelligence attack in elections. I know the secretaries of state generally are concerned about that."

He said his office will be working to spread more awareness of threats posed by AI in elections as well as talking to voters about double-checking the things that they see and not taking suspicious items at face value.

These sorts of AI robo calls and deep fakes are only becoming more prevalent, said Ashley Casovan, the managing director of the AI Governance Center at the International Association of Privacy Professionals. Now even the average person, like Huntley, is getting insight into the potential harms coming from these systems.

“There’s always been a concern about these technologies and how they’re used,” Casovan said. “But when people pick up the phone and they are getting these robo calls, it just makes it more real.” Besides the New Hampshire primary, the technology has already been used in elections throughout the country and across the globe, like a recent election in Slovakia.

Spreading awareness about this issue is great, she said, but her organization hopes for more regulatory actions and legislation as well.

NH lawmakers introduce bipartisan legislation to combat AI interference in elections

Rep. Angela Brannan, D-Concord, is the prime sponsor of HB 1596, which would require a disclosure of artificial intelligence usage in political advertising.

The bill would prohibit the use of deepfakes or deceptive artificial intelligence within 90 days of an election unless there is full disclosure.

The disclosure would say, "This [audio, image, or video] has been manipulated or generated by artificial intelligence technology and depicts speech or conduct that did not occur." If an audio like the Biden robocall, the disclosure would play at the beginning and the end, as well as within the audio if longer than two minutes.

“I think voters deserve accurate information, and it's important to recognize that AI is a rapidly advancing technology, making it super easy to deceive voters,” said Brennan. “Voters deserve better than that.”

First Amendment concerns have been raised in other states that have tried to prohibit deep fakes. Brennan said this bill includes a lot of safe harbors for these concerns. For example, it would not apply to satire or parody.

“There are certainly additional actions that could be taken and additional regulations that could happen and there are states who have tried going further,” Brennan said. “I'm just trying to take a small step forward to protect voters from being deceived from technology that really can be done in five minutes through an app on your phone.”

The bill is a bipartisan effort, with Republican sponsors like Rep. Ross Berry, R-Manchester. Brennan is hopeful the bill, which is being heard Tuesday in the Election Law Committee, will pass the House.

If it becomes law, it will go into effect on Aug. 1, well before the presidential election in November. Brennan said this was intentional.

“Because as we've already seen, in our first in the nation primary, this is technology that is accessible now, and it's going to be influencing our elections now. It's already happened. And if we don't start taking steps to try to prevent this type of deception especially during this critical election year, I am afraid that voters will be deceived,” she said. “They deserve much better than these kinds of dirty tricks.”

AI interventions can “erode senses of democracy”

Casovan says one of the significant harms AI poses in election interference is the erosion of trust and the feeling of democracy in society.

Indeed, these were Huntley's fears after receiving the call.

“You just don’t know how far they’re going to take it. It’s a clear example of voter suppression, they’re just desperate to take away our liberties and democracy in doing this,” Huntley said. “It was just horrible.”

In a letter to the Department of Justice and the Cybersecurity and Infrastructure Security Agency sent Tuesday, Sen. Jeanne Shaheen, D-New Hampshire, urged the departments to "address the threat that AI-generated deep fake technologies pose to the democratic process."

"I am convinced the proliferation of disinformation and misinformation about our elections and political system pose one of the gravest threats facing the United States today," Shaheen wrote. "I trust that the DOJ and CISA share this concern, and I look forward to hearing how the two agencies plan to address this critical threat to our democracy."

While the International Association of Privacy Professionals doesn’t advocate for certain types of policy reform, Casovan said that they are definitely in support of efforts like content labeling that the New Hampshire bill proposes.

“There's more prolific use of AI than we've ever seen in every single industry, that are doing things like interfering with basic tenets of society,” said Casovan. “We're really heartened by all of the legislation in various different forms that's coming and I think can and will play a really important role is to stop some of this behavior.”

This article originally appeared on Portsmouth Herald: NH bill targets AI in elections after Biden deepfake in primary