The ACLU is predictably worried about the use of facial recognition by FBI, and it needs to oblige the agency to reveal its practices. It just filed a court case in opposition to the Justice Department, the FBI, and the DEA asking them to hand over records displaying “where, when, and how” they employ facial recognition technology. The civil liberties group was worried that these systems can “basically change” society and result in steady surveillance, and highlighted to the public stances and FBI’s history as reasons to be worried.
The FBI has been involved in “political policing,” the ACLU claimed, comprising spying on peaceful protesters. That lifted the ability for abuse in opposition to innocent targets. The agency also stated that it did not require to show probable cause to employ facial recognition, and could not verify if it respected “constitutional bans” to tell defendants in criminal offenses when the tech was comprised. And these issues assume the systems are precise, which they sometimes are not—the ACLU pointed to researches displaying gender and racial biases in facial recognition.
This is not the first time the ACLU has asked for transparency. It filed requests of Freedom of Information Act with the DEA and FBI in January. Both bodies recognized the requests, but did not offer any “responsive papers,” the ACLU claimed.
On a related note, Facebook has already ended by default employing facial recognition, but now it may have a method for users to dodge facial recognition in total. Its scientists have designed an AI system that can “de-recognize” users in real time, comprising live videos. The method connects a trained face classifier and an adversarial auto-encoder to ever-so-slightly distort face of a person in a manner that puzzles facial recognition networks while maintaining a natural look that remains recognizable by honest-to-goodness people.