AI Raids Fooled Millions — DHS Fires Back

Woman with glasses shouting at a rally — Photo: Grossinger / Shutterstock

The most incendiary “ICE riot” visuals that helped frame recent unrest in U.S. cities were not what they claimed to be at all; a substantial share of the viral material was recycled from older protests, miscaptioned from other countries, or outright generated by artificial intelligence.

Key Points

Multiple fact-checking investigations show that a widely shared “ICE riot” clip predates the protests it was said to depict and is unrelated to the alleged ICE killing it was tied to.
Other supposed attack-on-ICE videos were traced to events in Mexico or to older U.S. protest footage, while still more “raid” clips turned out to be AI-generated fabrications.
The Department of Homeland Security has publicly catalogued false narratives targeting ICE, including debunking claims of dramatic Los Angeles operations that never occurred.
Independent reporting finds that anti-ICE protests themselves are real and largely grassroots, but the visual story told online is heavily polluted by fake or misattributed content.

How the “viral ICE riot” story fell apart

The starting point for many people’s understanding of the recent ICE-related unrest was not a police report or a news story but a short, punchy video on social media: burning streets, crowds clashing with officers, a sense that the country had tipped into chaos. That clip was widely described as proof that a woman’s killing by an ICE officer in Minnesota had triggered catastrophic riots in multiple U.S. cities. Only later did methodical verification work reveal a very different picture.

Misbar, a fact-checking outlet that specializes in tracing viral content, examined the video that was repeatedly captioned as “Minnesota ICE riots” and found that it did not depict Minnesota at all. The footage showed confrontations outside the Robert F. Kennedy Federal Building in downtown Los Angeles during anti-ICE demonstrations on June 9, 2025, and had first appeared on TikTok days later. In other words, the clip was real protest footage—but from another city, another year, and another context entirely. It had been retrofitted with a new story.

In parallel, AFP and other verification desks reviewed a second piece of viral content that purported to show Los Angeles residents physically attacking ICE agents. That video, too, spread under captions linking it to the Minnesota killing and to a supposed escalation of violence against immigration officers. AFP’s investigation traced the clip back to a February 2026 brawl between workers and local police in a wholesale market in San Luis Potosí, Mexico. The visual details—the uniforms, signage, and language—matched Mexican authorities, not U.S. federal agents. Once again, the events in the frame were real, but they had nothing to do with ICE or U.S. protests.

Other misattributed “riot” imagery turned out not to be connected to any ICE controversy at all. DW and local outlets identified a separate clip, circulated as evidence that an ICE agent had been struck during Chicago unrest, as a 2023 Illinois State Police chase involving carjacking suspects. The officer in question was a state trooper, and Illinois State Police later confirmed the footage was from an older criminal pursuit with no ICE involvement. The incident had simply been recaptioned to fit the moment’s narrative.

Recycled protest footage and the illusion of escalating chaos

Once fact-checkers began looking systematically, they found that the mislabeling was not an isolated error but part of a broader pattern. The Los Angeles Times reported that some of the most dramatic visuals shared as evidence that the city had become a “war zone” under anti-ICE protests—police cars ablaze, streets in apparent collapse—were taken from 2020 demonstrations after the murder of George Floyd, not from the current immigration protests. X posts using that footage were later removed, but not before they had reinforced the impression of sweeping violence.

Similarly, AFP and others have documented cases where fiery protest scenes from 2020 Minneapolis unrest were reintroduced in 2026 with captions tying them to ICE-related incidents in Minnesota. Reverse image searches traced those videos back to reporting from the initial George Floyd protests, demonstrating that only the text had changed, not the images themselves. For viewers encountering the clips fresh, the distinction was invisible; the recycled material looked like live coverage of a new meltdown.

This recycling dynamic matters because it multiplies the apparent scale and ferocity of unrest without any corresponding change on the ground. One genuine protest can seed dozens of misleading narratives once its images are re-captioned and juxtaposed with unrelated claims. Mashable, surveying the misinformation swirling around the Los Angeles ICE protests, described recycled footage as one of the core trends: old videos re-emerging to heighten outrage, “overshadowing” actual events with borrowed spectacle.

AI-generated ICE raids and synthetic confrontation

On top of the miscaptioned real-world video sits another layer entirely: protest and enforcement scenes that never happened. Starting in late 2025 and accelerating into 2026, a cluster of short clips went viral purporting to show dramatic ICE workplace raids—Walmart cashiers lined up beside a van, McDonald’s staff marched out mid-shift, a Pizza Hut crew detained, agents storming a classroom or ambushing a food-delivery rider. Each clip was presented as raw documentation of aggressive new tactics.

Multiple independent fact-checkers, including AFP, Reuters, Full Fact, and Snopes, analyzed those specific videos and reached the same conclusion: they are AI-generated fabrications. Investigators pointed to telltale markers—uniform clip lengths, repeated camera moves, nonsensical signage, and visible watermarks from tools such as OpenAI’s Sora and Google’s Veo—as well as motion artifacts and “robotic” body movements inconsistent with real footage. DW’s fact-check team extended this analysis to clips claiming ICE agents had appeared at California State University or a high school soccer match, likewise finding them to be synthetic creations masquerading as news.

The phenomenon is not limited to moving images. VERIFY, a U.S. outlet specializing in digital provenance, examined a widely shared still image of a young girl apparently being arrested by ICE officers and found it was generated with an AI image tool; the creator himself confirmed to Snopes that the picture was artificial. Another viral image showed ICE detaining a man in a “Latinos for Trump” shirt; reverse image search located the original 2018 photo from a meatpacking-plant raid, revealing that the political logo had been digitally superimposed onto a real person’s clothing.

France 24’s explainer on genuine versus fake ICE arrests underscores the scale of this synthetic swarm: alongside “very real ICE arrest videos,” the outlet noted, there is a “series of AI-generated ones” circulating in multiple languages and feeds. The result is a media environment in which audiences cannot safely assume that any given enforcement clip or protest confrontation is genuine—even when it looks convincingly rough and handheld.

DHS pushes back: institutional debunks and political skepticism

Recognizing the growing distortion, the Department of Homeland Security has taken the unusual step of publicly cataloguing and rebutting false claims involving ICE. In June 2025, DHS released a statement titled “DHS Debunks Fake News Media Narratives from June,” listing a series of widely circulated misstatements and clarifying the underlying facts. Among them was the claim that ICE had conducted operations at Douglas Park in Los Angeles targeting nannies and caregivers; DHS described that allegation as “unequivocally FALSE.”

A second DHS communication focused specifically on “fake news demonizing ICE officers” and set out to “set the record straight” on Los Angeles operations. The message was as much about tone as detail: federal officials argued that viral narratives were painting agents as cartoon villains, attributing to them actions that had not occurred, and inflaming public opinion on the basis of fabrications. That stance dovetails with independent fact-checking but inevitably lands in a contested political space.

Critics of ICE and of the Trump administration’s immigration policy are inclined to see DHS messaging as self-serving. The Guardian, for instance, described misinformation around the L.A. ICE protests as “catnip for rightwing agitators,” emphasizing that false stories were being weaponized to discredit genuine activism and to justify harsher enforcement. Both supporters and opponents of ICE therefore operate in a feedback loop where corrections from the agency may be accepted or dismissed largely on partisan grounds, regardless of their factual accuracy.

What was real: grassroots protests and contested enforcement tactics

It is important to separate the question of whether particular videos are genuine from the question of whether the protests themselves were. On that second point, the evidence is clear: large, sustained demonstrations against ICE operations did take place in Los Angeles, Minnesota, Chicago, and other cities. PBS, in a detailed fact-check of President Trump’s claim that anti-ICE protesters were “paid agitators and insurrectionists,” found no substantiated evidence that the protests were staged. Instead, researchers documented a broad grassroots movement, with local residents forming the bulk of participants and professional organizers playing the limited role they typically do in any mass protest.

The same PBS review concluded that social media posts purporting to prove protesters were “paid” relied on AI-generated videos, recycled conspiracy tropes, or unverified assertions. The claim that the movement was scripted by “professional troublemakers” was classified as false. In parallel, mainstream reporting from outlets such as the Los Angeles Times emphasized that while demonstrations were disruptive and sparked intense debate, they had not produced the level of death or structural destruction implied by the most viral clips.

At the same time, there is documented tension and occasional violence at the intersection of federal immigration enforcement and civic protest. France 24, FOX segments, and independent coverage describe real cases in which agents have used force, protesters have been injured, and operations have sparked legal challenges. The existence of fake or miscaptioned video does not erase those realities; it complicates how they are perceived, contested, and remembered.

How miscaptioned and synthetic media reshape public perception

When you put these strands together—recycled protest footage, misattributed foreign incidents, AI-generated raids, and institutionally contested narratives—the result is a media ecology in which the “ICE story” most people see bears only partial resemblance to the underlying facts. Emotionally potent clips circulate at algorithmic speed; they are shared by accounts with clear ideological goals; they provide visual confirmation of whatever a viewer already fears or believes. By the time forensic work establishes that a video predates the alleged incident or was filmed in another country, the damage to public understanding is done.

Verification teams and digital forensics labs now treat provenance analysis—geolocation, timestamp checking, reverse image search, watermark detection—as essential infrastructure for covering protest and enforcement stories. The opportunities for deeper truth-finding are clear: platform-native logs could anchor the exact upload and dissemination history of key clips; FOIA requests and court records could confirm or refute alleged incidents; sworn testimony from officers and journalists could lock down what was happening off-camera. But those slow tools are constantly racing a fast, emotionally charged misinformation machine.

For citizens trying to make sense of contested ICE stories, the practical implication is stark. A single viral clip, no matter how compelling, is no longer enough to justify strong conclusions about what happened. The evidence assembled across multiple investigations—Misbar on Minnesota, AFP on Mexico, DW on AI fabrications, PBS on paid-protester claims, the Los Angeles Times on recycled violence—points to a consistent reality: many of the most dramatic “ICE riot” visuals are simply not true in the way they are being presented.

The protests, the politics, and the policy fights are real. The viral story that crystallized them in a single, supposedly definitive ICE riot video is not. Understanding that gap is the first step toward a more honest conversation about immigration enforcement, public dissent, and the fragile trust between what we see on our screens and what is happening in the streets.