Law Enforcement Use of Facial Recognition

Facial recognition is a powerful investigative tool that, when used responsibly and with proper safeguards, enhances public safety without compromising civil liberties.

Background

Facial recognition technology has become one of the most widely discussed and scrutinized tools in modern law enforcement. Used primarily as a forensic aid, the technology enables investigators to compare images—such as those captured from crime scene surveillance footage—against databases of known individuals to generate potential leads.

It has proven valuable in solving crimes, identifying missing and exploited children, apprehending human traffickers, and supporting counterterrorism efforts. At the same time, the technology’s growing capabilities have prompted significant public debate about accuracy, potential bias, and the appropriate limits of its use.

Independent testing by the National Institute of Standards and Technology (NIST) has shown that the best-performing facial recognition algorithms achieve accuracy rates exceeding 99 percent with virtually undetectable demographic differences, and are far more accurate and less biased than unaided human recognition.

IBIA Position

IBIA supports the responsible use of facial recognition by law enforcement, subject to clear policies, proper training, and robust safeguards. Biometric technologies should never be subject to outright bans, which pose serious risks to public safety and national security.

IBIA’s key positions on this issue include:

Forensic Use Should Always Be Permitted

Facial recognition for forensic investigation—that is, using the technology to analyze evidence after a crime has been committed—should always be allowed. Banning facial recognition would preclude its use in forensic analysis, severely limiting law enforcement's ability to solve crimes, identify missing and abused children, apprehend human traffickers, and respond to mass casualty events such as bombings and mass shootings.

Law Enforcement Should Use Facial Recognition for Lead Generation, Not Positive Identification

Facial recognition in law enforcement is used to generate investigative leads, not to establish guilt. The technology produces a set of potential matches that must be supplemented by additional evidence sufficient to meet evidentiary standards. Most jurisdictions configure their systems to return either no result or multiple candidates to reinforce that results are leads, not affirmative indicators of guilt, and many use human examiners to review results as an additional check.

Real-Time Surveillance Requires Court Authorization

While forensic use should not be restricted, IBIA believes that real-time surveillance using facial recognition should be subject to court order—similar to a wiretap—with a specific legal purpose, a defined time period, and a defined geographic area.

Proper Safeguards and Training Are Essential

Law enforcement use of facial recognition should be accompanied by clear policies stating that results are for lead generation only; supervisory review of results; case-based and periodic independent audits; mandatory training for all users; strict access controls limited to authorized personnel; an established audit trail; and penalties for misuse.

Law Enforcement Should Use Top-Performing Algorithms That Address Bias Concerns

NIST testing has demonstrated that high-performing facial recognition algorithms exhibit virtually undetectable demographic differences across racial groups and sexes. These are the algorithms that government agencies should be using. Automated facial recognition is more accurate and less biased than unaided human recognition, and banning the technology would mean foregoing improvements to an existing system that relies on far less accurate human judgment and eyewitness testimony.

Bans and Moratoria Are Counterproductive

Bans and moratoria would not only weaken public safety but would also cede U.S. leadership in the development of this critical technology to other countries. Continued use, refinement, and independent testing are essential to driving further improvements in accuracy and fairness.