This past week, Fortune reported that U.S. retailers were using facial recognition software to target shoplifters.[1] The technology works by scanning the face of customers entering a store and seeking to match the photograph with a group of previously identified individuals. According to the article, this previously identified group is created by the security personnel employed by the store.[2] This article raises several fascinating questions: (1) who owns an individual’s face; (2) what other databases can be compared to the facial scans; and (3) how accurate is the the scanning technology.
Facebook and Google, among other online tech giants, have been using facial recognition software to “tag” individuals in photographs for several years now.[3] This allows these social media platforms to identify users is in a photograph to post on their profile. In addition, it allows the platform to gather information about friend groups to provide superior marketing information to its advertisers. The U.S. Department of Commerce created a working group to determine the answer of who owns the facial scan.[4] However, privacy groups dropped out of the working group after they were unable to get companies to agree to basic privacy controls. There is a question of whether taking continual photographs requires the consent of those photographed, which would likely render facial recognition software impractical.[5]
While relatively new to the U.S., this technology has been used in Europe for years. For instance, a music festival in the U.K. adopted this technology to scan the faces of concertgoers.[6] The police claimed that the the system was to be used “to find organized criminals who prey on festivalgoers who are often victims of theft.”[7] This use of facial scanning shows the potential of the technology: software can incorporate broader databases into the facial database to catch individuals who may have a warrant out for failing to pay a parking ticket.
A final question raised is the accuracy of the technology. One study found that the FBI’s database, which contains the most pictures but is also one of the least technologically advanced, can only provide the right person 80% of the time.[8] Facebook, on the other hand, claims that it’s algorithm depicts the same person 97.25% of the time which is almost equivalent to a human.[9] However, there are no good studies that currently depict the real time offline accuracy of facial recognition software.[10]
Walmart’s experiment only lasted a couple of months after it found that it did not have a good return on investment.[11] In addition, many other companies appear reticent to use the technology. While Congress or the Department of Commerce can hopefully one day find a workable solution to these legal questions, many companies are already scared about the privacy backlash that may occur by adopting these scanners.
[1] Jeff John Roberts, Walmart’s Use of Sci-fi Tech to Spot Shoplifters Raises Privacy Questions, Fortune (Nov. 9, 2015), http://fortune.com/2015/11/09/wal-mart-facial-recognition/.
[2] Id.
[3] Help Center: Tagging Photos, Facebook, https://www.facebook.com/help/122175507864081 (last visited Nov. 16, 2015).
[4] Robinson Meyer, Who Owns Your Face, The Atlantic (Jul. 2, 2015), http://www.theatlantic.com/technology/archive/2015/07/how-good-facial-recognition-technology-government-regulation/397289/.
[5] Id.
[6] Paul Gallagher, Download Festival: Facial Recognition Technology Used at Event Could be Coming to Festivals Nationwide, Independent (London, U.K.) (Sept. 24, 2015), http://www.independent.co.uk/news/uk/crime/download-festival-facial-recognition-technology-used-at-event-could-be-coming-to-festivals-10316922.html.
[7] Id.
[8] Tim Cushing, The FBI’s Facial Recognition Database Combines Lo-Res Photos with Zero Civil Liberties Considerations, techdirt (Apr. 15, 2014), https://www.techdirt.com/articles/20140414/16045126909/fbis-facial-recognition-database-combines-lo-res-photos-with-zero-civil-liberties-considerations.shtml.
[9] Meyer, supra note 4.
[10] Id.
[11] Roberts, supra note 1.