Pro-immigration activists use AI to unmask Trump federal agents

In the interview at PoliticoDominick Skinner explains that he and his team of volunteers identified at least 20 masked agents in viral videos. These police officers challenge people in broad daylight on the streets, in cars, at home, in administrative buildings, and even on workplaces.

ICE federated agents are unmasked by an AI

Images that have shocked a lot with these armed federal agents with concealed faces during the arrests of high school students, children, mothers and other state citizens.

And precisely, the American government justifies these masks by protection against harassment. The Trump administration explains that these agents only do their missions. But voices believe that this is a lack of transparency when Ice multiplies the expulsions deemed arbitrary and racist.

Obviously, the defenders of Ice condemn the action of Dominick Skinner. Tanya Roman, spokesperson for Ice, declared to Politico that “Mally informed activists and others like them are precisely the reason why Ice’s courageous men and women choose to wear masks at the start, and why they and their family are increasingly targeted and attacked”.

The irony is that Dominick Skinner uses AI, a technology also used by ICE with its mass monitoring objective. According to experts, this is a “Expeculment machine” Built from private data from American citizens.

As precise PoliticoDominick Skinner uses AI to estimate the missing parts of a partially masked face. Then, the image is subject to Pimeyes, a tool often denounced for the stalking it causes. Dominick Skinner explains that his volunteers are “Capable of reconstructing a face with AI as soon as at least 35 % of the face is visible”. The activist adds that he defends the “Purior Public” and the “Public responsibility”.

Except that here, specialists in the privacy questioned by Politico Do not believe too much about the reliability of this method. They recall that the facial recognition used by the police has already led to legal errors. Which could also be the case here.