A Netherlands-based immigration activist named Dominick Skinner is using AI and facial recognition to reveal the identities of masked US Immigration and Customs Enforcement (ICE) agents. Talk about turned tables — and a striking ethical paradox.
In an interview with Politico, Skinner claimed that he and his team of volunteers have so far been able to use AI to identify at least 20 ICE agents seen in video recordings that have gone viral of the masked figures arresting people — students, children, mothers, and American citizens included — in broad daylight. The videos are deeply troubling, in part because of the dystopian imagery of armed federal agents shielding their faces as they arrest people in streets, their cars, homes, government offices, and workplaces.
The federal government continues to claim that the masks serve to protect ICE agents from undue harassment for simply doing their jobs. Many, though, have condemned the effort as a symbol of ICE’s sweeping powers, and the lack of accountability or transparency that the agency is currently enjoying in its aggressive push to expand deportation efforts.
ICE and its defenders are foreseeably furious about Skinner’s use of AI to de-mask agents, with ICE spokesperson Tanya Roman telling Politico that “misinformed activists and others like them are the very reason the brave men and women of ICE choose to wear masks in the first place, and why they, and their families, are increasingly being targeted and assaulted.”
But as ICE itself continues to infuse AI into its ever-expanding surveillance apparatus — which includes a large effort to build what experts have called a “deportation machine” using private data of Americans — Skinner’s operation raises troubling questions about the current state of what’s possible with facial recognition tech, while turning the conventional power structure of surveillance on its head.
Per Politico, Skinner’s project works by first using AI to guess what the rest of an ICE agent’s face likely looks like, and then running that face through PimEyes, an off-the-shelf facial recognition model that’s long been mired in controversy around its training practices, ethics, and use for dangerous behavior like stalking.
Skinner and his team, he claims, are “able to reveal a face using AI, if they have 35 percent or more of the face visible,” he told Politico. He added that he doesn’t believe in “public justice,” but does believe in “public shaming and public accountability.”
Privacy experts that Politico spoke to had misgivings about how well facial recognition actually works on artificial images like the ones Skinner and his team are creating. That’s a fair point, given how frequently facial recognition utilized by law enforcement winds up linked to flimsy policing and wrongful arrests.
Skinner’s use of AI to publicly identify masked agents should raise alarm bells for other groups — for example, masked protestors fearing retaliation in a political environment increasingly hostile to First Amendment rights. And, well, pretty much anyone else whose face has at any point been online.
Videos of masked agents in American streets — not to mention American cities full of armed troops and federal agents — are surely unsettling, as are alarmingly easy-to-access facial recognition tools. Skinner’s activities may do well to serve as a reminder that, mask or no mask, AI and facial recognition have collided with our digital lives in a way that makes it next to impossible to be genuinely anonymous.
More on facial recognition: Police Arrest Man With Almost Zero Resemblance to Actual Perpetrator Because AI Told Them To