Many West Virginia residents carry smartphones that can be unlocked with a glance. Facial recognition technology is now extremely commonplace, but it is not always reliable. The facial recognition systems on phones and other devices work well because they perform a single and very simple task. They compare a face with a stored image. The limits of the technology become more apparent when millions of images are analyzed in search of a match.
When police departments around the country have photographs of a suspect that they would like to identify, they often call a New York-based company that offers facial recognition analysis and maintains a database of more than 30 billion images scraped from the internet. Law enforcement has accessed this database more than a million times, but the leads it provided did not always send investigations in the right direction. More rigorous standards were introduced in Detroit after facial recognition misidentifications led to false arrests and raised questions about the reliability of the technology.
The federal government
The facial recognition technology that police departments across America use every day is banned in Canada, Australia and the European Union. A ban in the United States is unlikely because the technology is used widely by the federal government. U.S. Citizenship and Immigration Services recently unveiled a facial recognition system that will be used to keep track of asylum seekers. The technology that was once reserved for law enforcement is now being shared with criminal defense attorneys. The company that maintains the database changed its policy after its technology proved that an accused individual was innocent.
Identifications are unreliable
Identifications are unreliable when they are made by people or machines. Misidentifications made by eyewitness are the leading factor in wrongful convictions according to the Innocence Project, and a string of high-profile false arrests linked to facial recognition services suggest that the latest technology is just as unreliable.