Footage of real ICE raids are dramatic and disturbing enough as it is. Now, Facebook users are being duped by AI-generated videos that portray even more extreme deportations happening in public spaces, from city streets to Walmarts.
As 404 Media reported Wednesday, a number of deepfake clips are going viral on the Meta-owned platform, which changed its content moderation rules after the 2024 election to have fewer restrictions on political topics like immigration and gender.
One viral clip purports to show Latino Walmart workers being lined up and put into a van with “IMMIGRATION AND CERS” written on the back. One employee is ordered to step up by an ICE agent with a “POICE” badge, but then does a weird, sideways walk back into the line, suggesting something’s off.
Other videos posted by the same Facebook account ― “USA Journey 897” ― show McDonalds employees being similarly targeted in immigration raids or food delivery workers on bikes being ambushed by ICE agents.
None of these videos are real. They’re riddled with telltale AI signs ― in the Walmart video, the man’s unnatural gait and the misspellings throughout give it away ― but people are still falling for them.
“It’s about time they go to Walmart,” one man commented on the clip, which has 4.1 million views, 16,700 likes, 1,900 comments and 2,200 shares on Facebook.
“I almost believed that up until the last guy in the video lol,” another man said of the weird walker. (Admittedly, the clip, likely generated by OpenAI’s Sora, is more sophisticated than the AI we’ve seen in the past.)
Another glossier, more dramatic AI-generated video shows a young woman being deported while screaming for the baby she’s leaving behind.
Judging by the comment section, more people fell for this one. There’a few comments along the lines of “How can society tolerate this?” but mostly people applauded what they saw. “Love ICE ! Get ’em all!” one woman wrote with a heart emoji.
As 404’s Jason Koebler wrote, it’s unnerving that an account is spreading such disturbing, dehumanizing videos of immigration enforcement and that Facebook’s algorithm is rewarding whoever’s behind it with views and likes.
What’s equally disturbing is how many people are taking the bait. That’s in part because deepfakes are getting more realistic-looking, said Sherry Pagoto, a psychologist and director of the UConn Center for mHealth and Social Media.
She’s not surprised the videos are going viral. Content that triggers strong emotions — as deepfakes often do — generate tons of engagement, Pagoto said. In a system where clicks translate directly into revenue, the people making these videos know exactly how to game it: An AI-generated clip that taps into polarizing politics or harmful social tropes is almost guaranteed to do serious numbers.
Just last week, racist AI-generated videos purporting to show Black SNAP recipients having their EBT cards declined or claiming to sell their SNAP benefits for profit went viral on social media.
“Once our emotions take over, whether it be fear or rage, our ability to scrutinize the veracity of the content is weakened,” she said. “This is why misinformation is often designed to elicit strong emotion.”
A quick critical look tends to reveal the obvious signs of AI ― wonky hands, misspellings, audio discrepancies, a general “uncanny valley” feeling ― but not everyone is looking so closely.
“If you are giving something half your attention as you’re sitting in front of the TV or waiting in line for something, if AI is rendered well enough, it could look real,” said Jen Golbeck, a professor at the College of Information at the University of Maryland who studies AI and social media use.
Obviously, none of this bodes well for upcoming elections. Given the gullibility of Facebook users ― and improvements in AI ― bad actors will likely increase their use of AI-generated videos to tap into political-social discourse, said Cayce Myers, a professor and director of graduate studies at Virginia Tech School of Communication.
We’ve seen Russia mount concerted campaigns to spread disinformation during our elections and sow division in the U.S. But now, even political campaigners are playing around with deepfakes; before last month’s election, Senate Republicans posted a 30-second video on X and YouTube taking Sen. Chuck Schumer’s real words and using AI to invent fake video footage.
The fake footage of Schumer included a small transparent watermark of the National Republican Senatorial Committee’s logo on it and the words “AI generated” in the bottom right-hand corner. But if you were scrolling quickly through your feed, you could easily miss that.
“AI is such an inexpensive way to create political content and ads that resonate with audiences,” Myers said. “Even when these videos are obviously fake, they still fuel a larger political narrative.”
There’s absolutely a risk, especially among lower-information voters and less tech-savvy voters, that these videos could swing their opinion during big elections, especially if the videos confirm their existing biases, Golbeck said.
Golbeck pointed to how the Trump campaign, especially Vice President JD Vance, pushed the fabricated claim that immigrants in Ohio were eating people’s pets ― just one of the viral false narratives in the 2024 presidential election.
“The immigrant cat story was fact-checked and disproven, but it stuck,” she said. “Vance repeated the lie in the last month [of the election].”
What that (and these deepfake ICE and SNAP videos) reveal to Goldbeck is “that there is a large group of the electorate and politicians, elected officials, and legislators, who want to tell stories that give the vibe they want without any interest in the truth.”
“AI-generated videos will make it easier to create clips that hit that vibe,” she said. “I think we are in a timeline where the importance of truth in elections has already been lost.”
Technology may help create more disinformation and spread it more widely, Goldbeck said, but the real challenge is that it is now mainstream for high-level officials to create stories with provably false claims and commit to them, even in the face of overwhelming evidence.
“The voters who accept it do so without any concern for the truth, but in celebration of their feelings being validated,” she said.