From Facebook and TikTok to Instagram and X, AI-manipulated images and videos depicting Alex Pretti’s final moments have proliferated across the internet since his fatal shooting by federal officers in Minneapolis last weekend.
The rapid spread of media altered by artificial intelligence, much of which shows Pretti collapsing in the seconds after he was shot, has clouded key details of the shooting on social networks. Unlike other AI-generated deepfakes that portray entirely unrealistic scenes and are easily identified as fake, many of the AI-altered depictions of Pretti’s shooting appear to have been based on verified images, mirroring reality enough to confuse and mislead many online.
And even as awareness of the capabilities of advanced artificial intelligence spreads, some online are extending their skepticism to authentic media, falsely claiming that legitimate photos and videos of Pretti have been altered by AI.
One image that appears to have been manipulated with AI, showing the ICU nurse falling forward as a federal officer points a gun at his back, has been viewed over 9 million times on X (even as it received a community note that the image had been enhanced by AI). Among other AI-fabricated details, the still image features an ICE officer without a head.
Sen. Dick Durbin, D-Ill., displayed the image during a speech on the floor of the Senate on Wednesday, apparently without realizing the image was not authentic.
In a statement to NBC News, a spokesperson for Sen. Durbin wrote: “Our office used a photo on the Senate floor that had been widely circulated online. Staff didn’t realize until after the fact that the image had been slightly edited and regret that this mistake occurred.”
Other posts have featured realistic video, including an AI-generated video on TikTok of Pretti talking with an ICE officer and one on Facebook of a police officer accidentally firing Pretti’s gun. The Facebook post, which has been labeled with a community note as being “AI-enhanced,” has more than 44 million views. It has not been established whether that officer fired a shot with Pretti’s gun.
Ben Colman, co-founder and CEO of deepfake-detecting company Reality Defender, said the rampant saturation of posts with AI-involved media of the shooting was worrisome but unsurprising. AI-modified images attempting to unmask the ICE officer who fatally shot Renée Nicole Good, the other U.S. citizen killed in Minneapolis in recent weeks, began circulating the internet earlier this month, leading many online to inaccurately identify other people as the agent.
“Over the last couple of months, we’ve seen on social media a sizable uptick of photos that contain AI-generated ‘enhancements’ of grainy, blurry photos,” Colman told NBC News. “The problem with these deepfakes — and they are still deepfakes in the end — is that they are crude approximations at best, complete fabrications at worst and do not accurately enhance or unmask individuals in the end.”
“Details like the missing head of an individual in the photo show just how damaging it is for these fake photos to go viral,” said Colman, referring to the photo shared by Durbin in the Senate. “If we take fake photos as real regardless of the context, it distracts and can detract from truth and fact.”
The spread of AI media has also led many online to mistakenly claim that genuine videos of Pretti are inauthentic. Experts have worried that this dynamic could allow a phenomenon referred to as the liar’s dividend, in which bad actors use the claim that authentic media is AI-generated in order to create distrust and avoid accountability.
Three videos independently verified by NBC News show Pretti getting into an altercation with federal immigration agents in Minneapolis just over a week before his death. But one of the videos, shot by reporters with The News Movement, is being labeled as AI-generated by some social media users.
The video showing Pretti kicking the back of a vehicle used by agents before being wrestled to the ground was verified by his family through a representative. A witness who recorded a second verified video showing the same incident told NBC News that he hugged Pretti after the altercation with federal agents and asked if he was OK.
There are few, if any, tools available to news consumers that can accurately tell if a piece of content has been created or manipulated by AI. On X, the platform’s AI assistant Grok responded to inquiries into the authenticity of the footage, with several of its replies claiming the genuine video “appears to be AI-generated or altered.”
The spate of AI deepfakes has added to swirling misinformation surrounding Pretti’s shooting, as several right-wing influencers misidentified Pretti as another Minneapolis resident.
Surges in AI-mediated misinformation and disinformation surrounding breaking news have become more common over the past year, as advancing AI systems are increasingly capable of creating high-quality images and videos that blur the line between reality and fiction.
