By Rep. Patrick N. Long
The Federal Communications Commission (FCC) recently announced it is considering a petition from the right-wing Club for Growth to loosen restrictions on unsolicited political robocalls— including those using AI-generated voices—ahead of the Nov. 3 elections. You might remember this issue from the 2024 presidential primary, when an individual commissioned fake calls that were sent to thousands of voters using an AI-generated voice pretending to be Joe Biden. The incident resulted in criminal charges, including voter suppression and impersonating a candidate.
This new position from Trump’s FCC, which could open the door to even greater violations of public trust, poses a threat to our democracy. We need our New Hampshire election officials to step up and protect the integrity of our electoral process.
AI may give out false information about voting—like it did in the “Biden” case. One study found AI chatbots gave wrong or unclear information about voting 29% of the time, and even more with Spanish-language requests. Granite Staters are especially at risk because New Hampshire election laws have changed since 2024, so misinformation could gain credibility by quoting outdated “facts” from official sources. Some of this could be pre-empted with simple solutions such as official sources listing the date information is published, so it is clear which information is most recent.
The month after the “Biden” robocall, I attended a demonstration by a group called CivAI that illustrated just how easy it is to generate fake video and audio of anyone who has been in the news enough. You only need a few real images and audio clips to mimic them convincingly, then AI can do the rest.
The technology has improved even further since then. Even where we have laws on the books, loopholes, poor enforcement, the international scope of some threats, and the time-sensitive nature of elections mean that protecting our elections requires proactive security, not just prosecution later.
As reported by Granite Post on Oct. 5, Senate Republicans recently launched an AI deepfake attack on Leonard Bell, a candidate for state senate in Manchester. They edited Bell’s family out of a photo, changed his clothes, and added in a gold chain, pile of cash, gold bracelet, jewel-encrusted watch, and gold and diamond ring—conjuring racist stereotypes to portray Bell, a Black man, as a gangster.
New Hampshire News has regularly run such fakes, creating and publishing AI images of Chris Pappas in an Ozempic ad, and another video of Democratic Party chair Ray Buckley pushing him on a swing, with vaguely romantic implications. They’ve also created AI deepfakes of Jeanne Shaheen enjoying an expensive international vacation, and Stefany Shaheen on Epstein Island.
Every one of these was a felony under existing laws if the use of AI wasn’t disclosed. The law has not been enforced in most of these cases. It was inadequately enforced in the Biden deepfake case, where the perpetrator was prosecuted but successfully argued that because he thought the NH primary was a fake election, the law didn’t apply. Expect more convincing attacks in the future.
These desperate attacks on opponents are not the worst possible uses of AI in elections. Bad actors can also use it to suppress votes—fake instructions about where, when, or how to vote, fake instructions about which documents are necessary to vote or to register, fake video of a shooting or an ICE raid at a polling place. Or they could generate fake but convincing “proof” of fraud, as Russian actors did in 2024 when they created fake video of ballots in Pennsylvania being destroyed.
Imagine, on Election Day, all of your neighbors receive a video of the town clerk—or Secretary Scanlan—saying your polling place is closed due to a fire. The “clerk” says voters will be given a chance to vote next week in a new location before the video cuts to footage of your polling place engulfed in flames. This technology exists. How will they know it’s fake?
In 2002, John E. Sununu’s campaign illegally jammed phone lines to disrupt Jeanne Shaheen’s “get out the vote” efforts, which may have cost her the election. In 2023, Charlie Kirk attempted to “break the phone lines” of the Secretary of State’s office. These attacks used old-fashioned methods—an autodialer in the Sununu case and human callers by Kirk. With modern AI, attackers could generate an unlimited number of human-sounding phone calls. They could tie up a phone line, and a worker, for far longer than the autodialer, at a scale that would be virtually impossible with human callers.
Some clerks I have spoken to are already having trouble getting timely responses from the Elections Division. A flood of AI calls would completely incapacitate the Elections Division. The Secretary needs to establish a procedure for screening out AI-generated calls, giving clerks and moderators a means of reaching his office that can’t be flooded by AI, and routing calls that don’t have to be handled by his office on Election Day somewhere else—the attorney general’s office, a call center, the clerks—whichever is most appropriate.
AI-enabled attacks could flip a close election. We need proactive, aggressive education and security measures to deal with these threats. Secretary Scanlan is either unequipped or unwilling to be our strong advocate on this issue—it’s time we stop giving him a free pass and demand he do better before the situation gets even worse.
Patrick N. Long, State Representative for Ward 7 of Manchester, can be reached at 373 Hayward St, Apt 3, Manchester, NH 03103, 603-316-6938.
