Up Against the Deep Fakes
Opposing harmful disinformation shouldn’t be an issue with two sides, but MAGA Republicans are making it one.
STEVEN KRAMER SAYS HE WAS JUST doing his civic duty by pointing out how easy it is to distort reality and corrupt the political process through the use of artificial intelligence.
“Maybe I’m a villain today, but I think in the end we get a better country and better democracy because of what I’ve done, deliberately,” Kramer, a political consultant responsible for a robo call that convincingly mimicked Joe Biden’s voice, told the Associated Press in February. He said the call, which went to thousands of New Hampshire residents days before that state’s first-in-the-nation Democratic primary on January 23, was meant as “a way to wake up the whole country” to this looming threat.
The robocall’s AI-generated version of Biden’s voice urged voters not to vote for him in the primary but rather “save your vote for the November election,” because they could only vote in one or the other. It warned his supporters that voting for him in the primary “only enables the Republicans in their quest to elect Donald Trump again.” For added authenticity, the voice used Biden’s signature homespun expression, “What a bunch of malarkey.”
That’s also pretty much what authorities think of Kramer’s claim that he was doing his nation a solid by hiring a New Orleans street magician to make this fraudulent recording. Kramer is now facing a $6 million fine from the Federal Communications Commission as well as thirteen felony and thirteen misdemeanor criminal charges in the state of New Hampshire. The felony counts are for violating a state law against using misleading information in an attempt to deter someone from voting; the misdemeanor charges are for falsely impersonating a candidate. He is scheduled to appear in court on June 5.
Kramer’s robocall stunt is said to be the first time that “deepfake” technology has been used in a U.S. election. It surely will not be the last.
“A variety of threat actors will likely attempt to use generative artificial intelligence (AI)-augmented media to influence and sow discord during the 2024 U.S. election cycle,” states an analysis compiled by the Department of Homeland Security and shared with CBS News. “As the 2024 election cycle progresses, generative AI tools likely provide both domestic and foreign threat actors with enhanced opportunities for interference by aggravating emergent events, disrupting election processes, or attacking election infrastructure.”
But just as this new technology intensifies the threat posed to our democratic institutions, Republicans are poised to kneecap one of the agencies that could push back against the expected tsunami of election-related falsehoods.
The GOP’s knives are out for the Cybersecurity and Infrastructure Security Agency (CISA), part of the U.S. Department of Homeland Security, whose mission includes “secur[ing] both the physical security and cybersecurity of the systems and assets that support the nation’s elections” as well as “address[ing] common disinformation narratives by providing accurate information related to elections.”
CISA, according to House Judiciary Committee Chair Jim Jordan (R-Ohio), has “directly and through proxies, censored Americans’ constitutionally protected speech.” The House’s Select Subcommittee on the Weaponization of the Federal Government, which Jordan also chairs, has raged against CISA for having “metastasized into the nerve center of the federal government’s domestic surveillance and censorship operations on social media.”
Meanwhile, the Heritage Foundation’s voluminous Project 2025 report, a blueprint for a Trump 2.0 administration, claims that the agency has been “weaponized to censor speech and affect elections.” It calls for CISA to be moved from the Department of Homeland Security to the Department of Transportation. (The DHS itself would be dissolved as part of the plan.) The report declares: “Of the utmost urgency is immediately ending CISA’s counter-mis/disinformation efforts. The federal government cannot be the arbiter of truth.”
CISA, IRONICALLY ENOUGH, WAS CREATED in 2018 during Trump’s term in office, but became one of the many targets of his boundless ire in 2020. That’s when Chris Krebs, then CISA director, pushed back against Trump’s claims of a rigged election—or, as he put it on what was then still known as Twitter, “a RIGGED ELECTION!”
Not so, countered Krebs, in a tweet of his own, on November 12 of that year: “America, we have confidence in the security of your vote, you should, too.” He also posted a link to a statement in which the members of two governmental councils on election infrastructure called the November 3 election “the most secure in American history.” Trump responded by tweet-firing Krebs, calling his comments “highly inaccurate.”
Democratic members of Congress see the writing on the wall. As Sen. Mark Warner, who chairs the Senate Intelligence Committee, told Politico, “Trump literally fired his director of CISA for accurately noting that the election of 2020 was secure, so of course one has to worry what he might try to do to undermine the agency’s critical work in a theoretical second term.” Added Rep. Glenn Ivey (D-Maryland), “There’s a good chance that he’d try to either eliminate it or totally neutralize it.”
The president, notes Politico, “cannot unilaterally kill CISA,” but he “could pack it with loyalists who would slash its operations. He could also whip lawmakers to wipe out the agency’s budget. And almost certainly, he would shutter its efforts to combat foreign disinformation.”
Much of the MAGA crowd’s animosity toward CISA stems from its efforts to notify social media platforms about posts that are believed to contain often dangerous disinformation.
A Trump-appointed federal judge in Louisiana and the conservative Fifth Circuit court of appeals ruled that these notifications ran afoul of the First Amendment, even though the platforms are free to disregard this input. The case is now before the U.S. Supreme Court, with a decision expected by the end of June. At oral argument in March, a majority of the court, including the conservative Justice Brett Kavanaugh and Chief Justice John Roberts, seemed to agree that these notifications were permissible.
But the pressure the right has brought to bear on government agencies dealing with the disinfo problem has had its intended effect. As the New York Times has reported, officials within the Department of Homeland Security and the State Department have “suspended virtually all cooperation with the social media platforms to address posts that originate in the United States.”
DESPITE THIS RETREAT, CISA’s CONTINUED commitment to caring about the truth puts it on a collision course with Trump and his minions.
The agency website still includes a section labeled “Election Security Rumor vs. Reality.” It declares that election officials throughout the land “work year-round to prepare for and administer elections, implementing a wide range of security measures” to ensure the integrity of the results. But, it warns, the fact that individual election jurisdictions do things differently “can lead to uncertainty in the minds of voters . . . that can be exploited by malicious actors.”
And how.
The site’s frequently asked questions section includes this inquiry: “Why do election results change overnight?” It explains that a careful and conscientious tally of the votes is a process that takes more than a few hours, thanks to late-arriving absentee ballots, the curing of ballots with minor deficiencies, and other typical challenges for election workers. CISA states:
All of these things together mean that election results will change in the days—and nights—after Election Day, sometimes significantly. It is not a sign of malicious activity, it is the opposite: it is a sign that election officials are diligently counting every eligible ballot to ensure they are included in the final tally of results.
Oh, the heresy.
Setting voters straight on how elections work and prosecuting those who craft fake robocalls to deceive voters are both parts of the same fight. In an age where virtually anything that can be imagined can be convincingly faked, we need all the truth defenders we can get, at the federal, state, and local levels.
According to the National Conference of State Legislatures, laws requiring disclosure when AI is used in campaign ads have now been passed in thirteen states: Alabama, Arizona, California, Florida, Idaho, Indiana, Michigan, Mississippi, New Mexico, Oregon, Utah, Washington, and Wisconsin. Two states, Minnesota and Texas, ban it altogether when used in the lead-up to an election to harm a candidate.
The disclosure rules are often surprisingly robust. In Mississippi, for instance, the ads must state “that the conduct or speech depicted did not actually occur.” The laws in six states provide for criminal penalties.
Legislation to require disclaimers with political ads that use AI has also been introduced at the federal level by Sen. Amy Klobuchar (D-Minn.) and Sen. Lisa Murkowski (R-Alaska). On May 15, it was placed on the Senate Legislative Calendar, meaning it is now eligible for floor consideration.
Further legislation regulating the use of generative AI in political campaigns is urgently needed. But the road forward on the deepfake issue could become very short if the deepest fake of all—Donald Trump—is not convincingly defeated this fall.
 | A guest post by
|