The Republican Party belongs to Donald Trump. He is the Big Boss. Cross him, and you will be punished. Just ask Kentucky Rep. Thomas Massie, who earned the president’s wrath by supporting the release of the Epstein files, opposing tax cuts for the rich in the “Big Beautiful Bill” and criticizing the president’s failed war of choice against Iran. In retaliation, Trump endorsed former Navy SEAL Ed Gallrein in the party’s May 19 primary.
Soon after, a pro-Trump super PAC called “MAGA Kentucky” targeted Massie with an artificial intelligence-generated video suggesting the Kentucky Republican was involved in a romantic “throuple” with Democratic Reps. Ilhan Omar of Minnesota and Alexandria Ocasio-Cortez of New York, both members of the progressive group informally known as “the Squad.”
“Thomas Massie, caught in a throuple in Washington,” the ad’s narrator says. “He’s cheating with the Squad on the America First movement. This is worse than adultery — it’s a complete and total betrayal of President Trump and Kentucky conservatives.”
Massie condemned the ad in a social media post: “It reeks of desperation, but they’re hoping the older generation won’t realize it’s an AI generated lie.” On Tuesday, he was defeated. Whatever role the ad played in the race’s outcome, AI-generated disinformation is sure to become a larger and more destructive force in American politics.
Since Trump’s return to power, we all have at some point asked ourselves “Did that really happen?” Or we’ve said, “That has to be fake,” only to, a few seconds later, think, “Actually, I’m not sure.”
Since Trump’s return to power, we all have at some point asked ourselves “Did that really happen?” Or we’ve said, “That has to be fake,” only to, a few seconds later, think, “Actually, I’m not sure.”
Does Trump really claim to be chosen by God — and that he is on a divine mission? Did he actually post an AI-generated image depicting himself as Jesus Christ? Did he share a fake video of himself as a fighter pilot dumping human waste on No Kings protesters? Did the president of the United States really dare to post disgusting AI-generated images of Barack and Michelle Obama portrayed as apes?
The answer, every time, is yes.
But before it can fully register, something worse arrives — the next day, or the next hour. The cycle doesn’t stop. The cognitive and emotional weight of it accumulates until something breaks. This is what it feels like to live in a society that has succumbed to what the late psychiatrist Robert Jay Lifton described as “malignant normality.”
In many ways, this is the issue that has come to define the Age of Trump, a time that is also being defined by the rise of AI.
Writing in the Guardian, philosophers Mark Alfano and Michał Klincewicz called this AI-generated “slopaganda” — and argued that it works precisely because the aesthetic presentation bypasses reasoning and hits our emotions directly.
Trump is not Christ, nor can he pilot an F-16 and drop waste on protesters. That’s not the point. The point is the feeling his creates, the paranoia that the average person is powerless against such dark forces.
What makes slopaganda so difficult to counter is its speed and reach. Once a false image takes hold, it is nearly impossible to uproot, and the flood of fabricated content has a corrosive effect: It makes us doubt everything. We may get better at spotting slopaganda, but in the process we also become more likely to mistake real things for fake things. The result is a legitimacy crisis for the institutions that hold democracy and society together.
The GOP has taken the lead in the creation and proliferation of slopaganda. In April 2023, the Republican National Committee previewed the harm that AI-generated videos would do to American politics when it shared a video created completely with AI-generated images predicting what would happen after a Joe Biden victory in 2024: hordes of brown people from Mexico overrunning American cities, China defeating the U.S. military and ruling the world, and American troops outside of a San Francisco closed because of crime and drugs. After a CNN town hall the following month, Trump himself hopped on the AI bandwagon by sharing a fake video of anchor Anderson Cooper declaring that Trump had “ripped us a new a*shole” at the event.
That same year, Darrell West sounded the alarm on this in an interview with Salon. A senior fellow at the Center for Technology Innovation of the Governance Studies program at the Brookings Institution, he warned that new AI tools were “democratiz[ing] disinformation and giv[ing] everyone the capabilities of troll farms.” The problem, West argued, was not just the volume of fabricated content but its sophistication; the tools had become so advanced that even experts struggled to distinguish the fake from the real. Because there are no legal guardrails limiting what candidates or their supporters could distribute, West concluded, the conditions were in place for American democracy to be decided by “false claims and inaccurate beliefs.”
That was three years ago. Since then, the tools have only gotten faster, cheaper, more widely available — and harder to detect.
In November, Rep. Mike Collins, R-Ga., who is running in the state’s GOP Senate primary to take on Democratic Sen. Jon Ossoff, released an AI-generated video showing Ossoff pledging allegiance to Senate Minority Leader Chuck Schumer. Four months later, the National Republican Senatorial Committee shared a video made with AI of Texas lawmaker James Talarico, the Democratic candidate for one of the Lone Star State’s U.S. Senate seats, using his own voice to read his old tweets in front of an American flag. An anti-redistricting group in Virginia released a video ahead of the commonwealth’s April referendum on redistricting that depicted Democratic Gov. Abigail Spanberger — who was backing the ballot initiative — setting a barn on fire. A voiceover accused her and Democrats of wanting “to burn Virginia’s democracy to the ground.”
A 2024 Pew Research poll showed that 82% of Americans were concerned that AI would be used to create and spread false information about the candidates. Other research shows that a majority of Americans are not confident that they could distinguish between an AI-generated video and a real one.
A healthy democracy and a functioning society depend on a shared sense of what is real. When leaders and the public can’t even agree on basic facts, we lose the ability to solve important problems.
A healthy democracy and a functioning society depend on a shared sense of what is real. When leaders and the public can’t even agree on basic facts, we lose the ability to solve important problems.
Legislation has been proposed at both the state and federal level to curtail the use of AI because of the harm it is doing to democracy and society. On a practical level, the technology is advancing rapidly, and lawmaking takes time. Limiting the use of AI in political advertising and campaigns will also need to overcome objections about free speech.
As Steven Rosenbaum, author of the new book “The Future of Truth: How AI Reshapes Reality,” explained to POLITICO, these attempts to ban political deepfake are “well-intentioned, but ultimately they fail. The problem is these technology companies are incredibly agile . . . I don’t think we’re going to solve the problem by legislating away particular apps.”
Want more sharp takes on politics? Sign up for our free newsletter, Standing Room Only, written by Amanda Marcotte, now also a weekly show on YouTube or wherever you get your podcasts.
Congressional Democrats have supported legislation to counter political deepfakes, but that nascent effort has so far been largely ineffective. The states have enjoyed more success. California is one of thirty-one states that have passed laws to restrict the use of political deepfakes.
Many states have mandated disclosure laws, requiring a clear label on any ad created with AI. A smaller number of states, including Maryland and Massachusetts, have tried to ban the use of AI deepfakes under existing anti-fraud laws. The central problem with this approach is that the bans are typically for a specific window of time — 30 days, for example, in Texas — before the election, and in some cases, AI-generated deep fakes are permitted as long as they are accompanied by disclaimers. These bans have enjoyed bipartisan support on the state level, and have been enacted in both blue and red states.
Trump had planned to sign an executive order on Thursday that would have created a framework for the federal government to review national security dangers posed by advanced AI technology before it is released to the public. But only hours before the scheduled ceremony, he announced he was postponing the executive order because he did not want to take away America’s competitive advantage over China and other countries.
As AI in political ads continues to proliferate, Vladimir Putin’s Russia offers the clearest preview of where this can lead. Hypernormalization — the condition in which the gap between official and lived reality becomes so vast that people simply stop believing anything is true — guts civic life and society from the inside. Nothing really matters anymore. Participation feels pointless. Why care? In response, many people seek out political strongmen and other authoritarian leaders and forces like Trump and MAGA politicians who promise action, community and simple answers to complex problems.
American democracy is collapsing rapidly. The attacks on common understandings of reality and truth and how this undermines the legitimacy of institutions and norms are pushing it over the edge.
Ultimately, hypernormalization is a long-term political and cultural problem that one election cycle cannot fix. The longer this state of hypernormalization and assaults on shared reality and truth lasts, the longer the recovery will take.
