The immigration crackdown ordered by US President Donald Trump has turned deadly, with the second American killed this month by Immigration and Customs Enforcement (ICE) agents.
Cellphone video from Minneapolis witnessSeveral ICE agents are shown slamming 37-year-old ICU nurse Alex Preti to the ground and then shooting him to death. He did not pull a gun, as the Department of Homeland Security initially said.Various cellphone recordings showed ICE officers removing guns from their belts before Preti was killed.
Earlier this month, ICE agents shot and killed 37-year-old Renee Gudas as she tried to drive away.
Videos shot from multiple angles also discredit official statementsGood allegedly tried to run the officer over and shot him at point-blank range by firing through the window.
Amid these incidents, social media is filled with a mix of real eyewitness videos and AI-generated fakes, complicating efforts to understand what actually happened. Las Vegas Metropolitan Police Department warnsIt said it had seen an increase in AI-generated images and videos related to the country’s military, adding that the country “does not participate in active immigration enforcement operations.” DW Fact Check investigated several viral clips.
Will ICE agents be arrested by police?
Claim: ICE officers are arrested and beaten by police, as seen in several videosPosting on different platforms and in different languages (here: hereand here).
DW Fact Check: fake
This can be seen from details such as garbled text in the video. Subway signs don’t make sense (‘Exit Ses’, ‘Sotreé Seet’, ’42eet’) and uniform logos are incorrect or misspelled (‘pice’, ‘IICE’).
For example, if a police officer grabs an ICE officer with his right hand, but his left arm hangs down with little movement, his body movements appear unnatural, exaggerated, or stiff. The dialogue looks jumbled, as if the AI forgot to add the responses from each of the other characters.
ICE agents behave like NPCs (Non-Playable Characters) in video games. — Background characters controlled by the game rather than the player: They don’t seem to react to what’s going on. Mouth movements when screaming appear abrupt and exaggerated.
Similar patterns can be seen in other AI-generated content from other demonstrators allegedly in conflict ICE officers.
Will ICE agents enter classrooms and college campuses?
Claim: ICE officers entered California State University looking for the student. Meanwhile, other agents showed up at a high school soccer game.
DW Fact Check: fake
Both videos published on social media platform TikTok are generated by AI. The logo of the AI generation software “Sora” appears in a video of a university classroom — This is proof that it was created with the help of AI and does not show actual real footage.
A video of an ICE officer staring into the crowd at a soccer game has a strange, glossy look to it. Searching for “Agreka Soccer” yields no results. The faces in the crowd appear distorted and the writing on the posters makes no sense. TikTok also added a warning that the video contains AI-generated content.
Are AI fakes drowning out real sighting footage?
“One of the problems with AI-generated content and fake videos circulating alongside real videos is that it becomes very difficult to tell what’s real,” said Courtney Radosh, director of the Center for Journalism and Freedom at the Open Market Institute.
The widespread availability of AI tools has made it possible for anyone to fabricate videos that look authentic.
Radosh warns that disinformation campaigns may intentionally release fake videos to drown out accurate records of deadly ICE encounters.
“The digital media literacy project teaches people how to spot misinformation and disinformation,” said Brittani Collard, associate director of MediaWise, Poynter’s digital media literacy project.He said when misinformation spreads, verified fact-checking tends to become invisible to many people.
“In fact, viral deepfakes can drown out real videos when it comes to algorithms, because more people are watching deepfakes,” she told DW.
It is becoming increasingly difficult to decipher what is true and what actually happened, which “could undermine the legal process, undermine trust in video evidence, and undermine the credibility of eyewitness testimony,” Radosh warns.
Is it possible to identify fakes generated by AI?
While it’s still possible, it’s becoming increasingly difficult as generation tools advance. Although there are detection tools like Hive Moderation,AI-generated tools evolve faster than the systems designed to identify them. Previous symptoms, such as unnatural blinking and distorted reflexes, occur less frequently.
Koller advises looking for clues in the video.
- Watermarks and AI tool identifiers?
- Weird phrasing, distorted text, or inconsistent lighting?
- Is there audio in the language you understand?
- Do you have a caption that seems sensational or lacks context?
- Is the report from reliable media?
- Is there additional footage or different camera angles?
- Verification of source account?
Radosh added: “When you’re trying to create a deepfake, it’s often virtually impossible to tell the real from the fake. Even sophisticated experts can’t always do it.”
She advocates for stronger technical protocols to authenticate authentic footage.
Why do we make these videos? Disinformation and profit
Experts say the motives vary, including malicious actors seeking to disrupt public debate and trolls seeking chaos.
Furthermore, the creation of AI content may also be motivated by economic benefits, according to Radosh. ICE raids can be a highly lucrative topic for gaining followers and increasing digital ad revenue.
At the end of the day, Radosh says, it doesn’t really matter who is behind the latest wave of AI content. An even more serious problem is the erosion of trust. “People are losing faith that facts can be proven.”
She says there are virtually no guardrails on how AI-generated videos can be monetized, and the social media ecosystem encourages disinformation for profit. And the results are very alarming. When people can’t tell the difference between a falsely created video and a real video, they often avoid the news altogether.
Rachel Baig and Ines Eisele contributed to this report.
Editor: Silja Thoms
