AI Soldier Videos Flood Social Media, Exploit Public Support
A wave of AI-generated videos depicting fake U.S. soldiers has swept across TikTok, Instagram, X, and YouTube, exploiting public support for American troops to push disinformation, generate revenue, and manipulate public opinion about the ongoing U.S.-Iran war. The phenomenon, detailed in a New York Times investigation published Tuesday, represents a convergence of accessible AI tools, platform economics, and wartime emotion that researchers say is outpacing society’s ability to respond.
The Many Faces of Fake Soldiers
The videos fall into several categories, each with distinct motives. Some depict crying soldiers expressing distress about being away from home — content designed to trigger an emotional response before critical thinking can intervene. Others, like the notorious “Jessica Foster” Instagram account, portray attractive female service members who are vocal Trump supporters, building massive followings before funneling users to paid adult content platforms.
As Straight Arrow News reported, these videos are engineered to exploit a psychological vulnerability. Sarah Barrington, an AI researcher at UC Berkeley, told the outlet: “We’re at a point where humans absolutely can’t tell the difference.” Her studies found that 60% of listeners could not distinguish real audio from AI-generated fakes, rising to 80% when a real voice was included alongside a generated one.
A Coordinated Influence Campaign
While some accounts appear to be individual profit-seekers, research by dotNex has identified a more troubling pattern: a coordinated influence campaign on TikTok involving more than 100 inauthentic accounts that posted over 52,000 AI-generated soldier videos, amassing 3.8 million views. The campaign, which researchers concluded is “likely of foreign origin,” appears designed to normalize the U.S. invasion of Iran and suppress anti-war sentiment.
“The objective is to create an illusion of public consensus” about the military operation, Luca Luceri, dotNex’s founder and CEO, told Straight Arrow News. The report found Chinese-language metadata in some accounts, suggesting “foreign, likely Chinese-linked automated infrastructure” may be involved, though definitive attribution remains elusive.
The Jessica Foster Case
The most high-profile example of the phenomenon was the “Jessica Foster” Instagram account, which amassed over 1 million followers before being exposed. The account depicted a blue-eyed woman in Army uniform who posed with President Donald Trump, Ukrainian President Volodymyr Zelensky, and Russian President Vladimir Putin — all generated by AI. The U.S. Army confirmed no record of anyone serving under the name Foster.
As The Washington Post reported, the account’s real purpose was to funnel users to paid adult content platforms. Sam Gregory, executive director of the video-advocacy group Witness, described Foster as “the apotheosis of what MAGA fantasizes about, all packed into one channel.”
Joan Donovan, an assistant professor at Boston University who studies media manipulation, warned of broader implications: “The danger of this is that we’re moving toward a society of the unreal.”
The Business Model Behind the Slop
Daniel Schiff, an associate professor at Purdue University and co-director of the Governance and Responsible AI Lab, explained the economics driving the phenomenon: “Content drives traffic, traffic drives money. These bad actors, the spreaders, are leveraging the same kinds of strategies that social media platforms built their own engagement methods on.”
Some accounts generate direct ad revenue from platforms based on follower counts. Others, like the Foster account, serve as billboards for external revenue streams. Schiff noted that operators could be making “hundreds or thousands of dollars, maybe in the tens of thousands” from these schemes.
Pro-Iran Propaganda in Lego Form
Separately, a network called “Explosive Media” has been producing AI-generated Lego-style propaganda videos for the Iranian government. As BBC News reported, these videos — viewed hundreds of millions of times — depict U.S. soldiers being chased by Iranian forces and Donald Trump falling through a whirlwind of Epstein files. The anonymous operator, who goes by “Mr Explosive,” admitted the Iranian government is a “customer.”
Dr. Emma Briant, a propaganda expert, said the term “slopaganda” is too weak to capture the power of these “highly sophisticated” clips.
Detection Tools Exist — But Go Unused
One of the most frustrating aspects of the crisis, researchers say, is that technical solutions already exist. The Coalition for Content Provenance and Authenticity (C2PA), formed by Meta, Google, Microsoft, and OpenAI, offers automatic detection and labeling of AI-generated content. But as Barrington noted, “most of these platforms don’t reliably use that very easy watermark.”
Even when platforms label content as AI-generated, research suggests labels may not prevent belief in false claims. And with nearly 40% of U.S. adults using TikTok — roughly half of whom turn to the platform for news — the potential for real-world harm is significant.
What Comes Next
The AI-generated soldier video phenomenon represents a convergence of multiple troubling trends: the democratization of powerful AI tools, platform business models that reward engagement over authenticity, the emotional vulnerability of audiences during wartime, and the exploitation of public support for U.S. troops.
New York has proposed “cigarette-style” warning labels on AI-generated content, and other jurisdictions may follow with stricter regulations. But as detection struggles to keep pace with generation, and as platforms fail to consistently enforce their own policies, the burden increasingly falls on individual users to exercise media literacy — a task for which research shows most are ill-equipped.
As Barrington put it, the best defense is basic: “Can you answer these basic questions about this thing you’re sharing?” But she stressed that the responsibility was never meant to fall on users alone. “We unleashed a lot of these capabilities before we had the detection capacities,” Schiff said. “We’re now reactively trying to address these issues.”
The question remains whether reactive measures can catch up to a technology that is evolving faster than society’s ability to govern it.