AI Case Study

"FBI’s Next Generation Identification system (NGI) is a massive biometric data- base that includes fingerprints, iris scans, and palm prints collected from millions of individuals, not just as part of an arrest, but also for non-criminal reasons like background checks, state licensing requirements, and immigration. e-Interstate Photo System (IPS) is the part of NGI that contains photographs searchable through face recognition. Each of the biometric identifiers in NGI is linked to personal, biographic, and identifying information, and, where possible, each file includes multiple biometric identifiers. "

However, as the Electronic Freedom Frontier (EFF) notes, there are several concerns with such a facial recognition database. "According to the GAO report and FBI’s responses to EFF’s Freedom of Information Act requests, FBI has conducted only very limited testing to ensure the accuracy of NGI’s face recognition capabilities. Further, it has not taken any steps to determine whether the face recognition systems of its external partners—states and other federal agencies—are sufficiently accurate to prevent innocent people from being identified as criminal suspects. FBI admits its system is inaccurate, noting in its Privacy Impact Assessment (PIA) for the IPS that it “may not be sufficiently reliable to accurately locate other photos of the same identity, resulting in an increased percentage of misidentifications.” However, FBI has disclaimed responsibility for accuracy in its face recognition system, stating that “[t]he candidate list is an investigative lead not an identification.” Because the system is designed to provide a ranked list of candidates, FBI has stated the IPS never actually makes a “positive identification,” and “therefore, there is no false positive rate.” In fact, FBI only ensures that “the candidate will be returned in the top 50 candidates” 85 percent of the time “when the true candidate exists in the gallery.” It is unclear what happens when the “true candidate” does not exist in the gallery, however." EFF notes that the median quality of images in the database is below the recommended resolution for facial analysis and that accuracy tends to decrease with the size of the database, given how many people in a population look similar.

Additionally, the FBI has a Facial Analysis, Comparison, and Evaluation (FACE) Services Unit which has access to a substantially larger database of over 400M images for facial recognition. "As of December 2017, FBI’s FACE Services Unit was conducting more than 7,000 searches per month".