The Trump administration has shared AI-generated and AI-altered images online, prompting misinformation experts to warn about the risks of blurring what is real and what is fake. The concern sharpened after an edited arrest image of civil rights attorney Nekima Levy Armstrong that showed her crying. Experts said the episode reflects broader problems as AI-edited imagery circulates widely on the internet.
Homeland Security Secretary Kristi Noem’s account posted the original image from Levy Armstrong’s arrest before the official White House account posted an altered version showing her crying. The doctored picture was described as part of a wider wave of AI-edited imagery shared across the political spectrum, the report said, since fatal shootings of Renee Good and Alex Pretti by U.S. Border Patrol officers in Minneapolis.
The report said misinformation experts are troubled by the administration’s use of artificial intelligence and fear it erodes public perception of truth and sows distrust. In response to criticism of the edited image, White House officials doubled down on the post, with deputy communications director Kaelan Dorr writing on X that the “memes will continue.” White House Deputy Press Secretary Abigail Jackson also shared a post mocking the criticism, the report said.
David Rand, a professor of information science at Cornell University, said calling the altered image a meme “certainly seems like an attempt to cast it as a joke or humorous post, like their prior cartoons. This presumably aims to shield them from criticism for posting manipulated media.” Rand added that he found the purpose of sharing the altered arrest image “much more ambiguous” than the cartoonish images the administration had shared in the past.
Zach Henry, a Republican communications consultant who founded Total Virality, said memes often rely on layered meaning that can be hard for outsiders to interpret. He said “People who are terminally online will see it and instantly recognize it as a meme,” while “Your grandparents may see it and not understand the meme, but because it looks real, it leads them to ask their kids or grandkids about it.” Henry also said a strong reaction can help the content go viral, and he generally praised the White House’s social media team.
Michael A. Spikes, a professor at Northwestern University and a news media literacy researcher, said the creation and dissemination of altered images, especially when shared by credible sources, “crystallizes an idea of what’s happening, instead of showing what is actually happening.” He said the government should be a place where people can trust information that is accurate “because they have a responsibility to do so,” adding that sharing this kind of content is “eroding the trust” and “really worries me a lot.” Spikes said he already saw “institutional crises” tied to distrust in news organizations and higher education, and that official behavior inflames those issues.
Ramesh Srinivasan, a professor at UCLA and host of the Utopias podcast, said people are questioning where they can turn for “trustable information.” He said “AI systems are only going to exacerbate, amplify and accelerate these problems of an absence of trust, an absence of even understanding what might be considered reality or truth or evidence.” Srinivasan said sharing AI-generated content invites everyday users to post similar material and gives permission to people in positions of credibility and power, including policymakers, to share unlabeled synthetic content.
The report also described an ongoing flood of AI-generated videos tied to Immigration and Customs Enforcement action, protests and interactions with citizens. It said that after Renee Good was shot by an ICE officer while she was in her car, AI-generated videos circulated showing women driving away from ICE officers who told them to stop. The report said fabricated videos have also circulated about immigration raids and about people confronting ICE officers, often yelling and throwing food.
Jeremy Carrasco, a media literacy content creator and debunker of viral AI videos, said most of these videos are likely produced by accounts he described as “engagement farming,” which he said aims to generate clicks using popular keywords, including “ICE.” Carrasco also said some people who oppose ICE and DHS may watch the videos as “fan fiction,” engaging in “wishful thinking” that they are seeing real pushback against the agencies and their officers. He said, however, that “most viewers can’t tell if what they’re watching is fake,” and he questioned whether they would know “what’s real or not when it actually matters, like when the stakes are a lot higher.”
The report said Carrasco believes many viewers may not recognize signs of AI generation even when errors are visible. Carrasco suggested that watermarking systems embedding information about a piece of media’s origin into its metadata layer could help, noting that the Coalition for Content Provenance and Authenticity has developed such a system. He said he does not think widespread adoption is likely for at least another year, and he added, “It’s going to be an issue forever now,” saying “I don’t think people understand how bad this is.”
The concern, experts said, is not limited to immigration enforcement content. The report said fabricated and misrepresented images following the capture of deposed Venezuelan leader Nicolás Maduro “exploded online earlier this month,” and that experts expect the spread of AI-generated political content to become more common. The Associated Press report was also credited to writers Jonathan J. Cooper in Phoenix and Barbara Ortutay in San Francisco.