Experts say social media feeds are being swamped by an unprecedented wave of AI-driven disinformation as the United States heads toward its November midterm elections. They warn that many of the false posts are likely "slipping through the cracks" just as tens of millions of Americans prepare to vote.
Social media companies have pulled back on protecting voters from election manipulation. After the Cambridge Analytica scandal, the backlash against Facebook pushed big platforms to strengthen their trust and safety operations. A counter-backlash from the right, driven by fears of speech restrictions, grew stronger after Elon Musk took over X and after Donald Trump's 2024 election win. The result has been significant layoffs on the relevant teams, while platforms lean more on AI than on human reviewers.
Frances Haugen, a former Facebook data engineer who became a whistleblower and transparency advocate, said the technology has changed the economics of influence work. "You can run an information operation now with quality that was impossible in 2016 for super, super cheap," she said. Because operatives can keep adjusting a post until it passes moderation, "some of them are going to get through".
Campaigns from Pennsylvania to Texas have been scrambling to answer lies that spread quickly online, and many are unsure where to turn for help. Meta and X did not respond to requests for comment. YouTube, owned by Google, said all its users are subject to the same guidelines.
In Texas, Democratic Senate candidate James Talarico took time off earlier this month to recover from the flu. Social media instead carried a false story that he was being treated at a mental health facility in Ohio. The claim was pushed by a rightwing provocateur who published a deceptively edited phone call with a hospital staffer. Talarico's campaign released security-camera video showing him at home in Texas, and he said he had never been to Ohio. The lie still circulated for several days.
The Digital Democracy Institute of the Americas tracked how the story mutated once conservative Spanish-language media picked it up, turning into a false claim that he had attempted suicide. Its researchers said such shifts make the original correction insufficient to address every version. Founder and executive director Roberta Braga said platforms enforce their rules more consistently in English than in Spanish, and that increased use of AI moderation means "even more is likely slipping through the cracks". Talarico's Republican opponent, Ken Paxton, also ran an advert earlier this year using AI-generated footage of Talarico supposedly rewriting the Bible. Talarico told NBC News in September that such deepfake ads add nothing to civic discourse.
In Pennsylvania, viral posts claimed an undocumented immigrant arrested in the central part of the state was a canvasser for Democrat Janelle Stelson's congressional campaign. An official Department of Homeland Security statement then amplified the claim, with ICE deputy director Charles Wall saying it was unbelievable that Stelson would employ an illegal alien. Her campaign released records showing the man, Chimaobim Adedayo Igbojekwe, held no paid position, although an initial police report said he had claimed to be a campaign worker. Republican congressman Scott Perry's campaign continues to run ads containing the claim. Stelson's team sent cease-and-desist letters to media outlets, a Fox News reporter removed a post on X, and Citizens for Responsibility and Ethics in Washington filed a complaint alleging a Hatch Act violation by DHS.
Nora Benavidez of Free Press said Trump "has made clear that he intends to interfere in the elections" and has used agencies across the federal government. Stelson's campaign has not yet complained directly to the platforms, but a spokesperson said it continues to consider every legal avenue against the false claims.
The major platforms do have rules, though they are narrow. Meta requires disclosure of AI content in election ads, labels AI images identified through C2PA metadata, and bans posts that misstate voting dates, locations or legal requirements. It does little moderation of political commentary, and Free Press noted it narrowed restrictions in 2023 to allow debate over past elections. YouTube also stopped removing false claims about past elections that year, and defines election misinformation narrowly. Google offers likeness detection and favors information panels and authoritative sources over takedowns. X replaced internal safety reviews with crowdsourced community notes, and experts say its synthetic media policy means little without enforcement.
Haugen argued that moderating individual posts is no longer enough when AI can keep tweaking content until it gets through. Attention, she said, must turn to the systems of distribution, including their biases and the feedback cycles that let communities dampen harmful content.