Ethical use of facial recognition technology


by Chris Becker   |   Michigan Bar Journal


Prosecutors understand that their primary role in the criminal justice system is to seek justice. This role is specifically set forth in the Michigan Rules of Professional Conduct 3.8, which states:

A prosecutor has the responsibility of a minister of justice and not simply that of an advocate. This responsibility carries with it specific obligations to see that the defendant is accorded procedural justice and that guilt is decided upon the basis of sufficient evidence.

Perhaps the most critical stage in exercising this responsibility occurs with the decision to authorize or deny criminal charges submitted by a police agency. The mere act of charging someone with a crime has the potential to alter the course of a person’s life. Therefore, a prosecutor’s decision to do so should never be made lightly.

Exercising discretion in charging decisions has been both improved and complicated by the use of technology. The prevalence of videos, mobile phones, and other technological advances has increased the amount of information and evidence available to prosecutors. One tool that has been publicized in recent years is the use of facial recognition technology.

As with any technology, understanding the limitations of facial recognition technology is essential to its ethical use. Because of that, both the Michigan State Police (MSP) and the Prosecuting Attorneys Association of Michigan (PAAM) have adopted best practices to help ensure this powerful and very useful tool is understood and used appropriately.


The MSP Biometrics and Identification Division includes the Statewide Network of Agency Photos (SNAP) Unit. SNAP has access to a large number of facial images, which it uses to assist in generating leads for law enforcement.

Facial recognition technology can be used in a variety of ways. For example, a police agency investigating a crime can submit a photo or video to SNAP, which then uses a software application to process the image and create a digital map of the subject’s face. The software compares that map with the images in its database to create a gallery of people who might be the person depicted in the submitted image (assuming it is of sufficient quality). An examiner then compares the submitted image with the computer-generated list of potential leads and determines whether there is a close enough similarity in any of the images to present them to the submitting agency as a possible lead. Should the examiner determine there is a viable lead or candidate, that decision is reviewed by another trained examiner before the information is submitted to the police agency.

Any SNAP Unit report to police agencies emphasizes that the leads are only that — they are not positive identification, they do not provide probable cause to arrest, and further investigation must take place. And because its method only generates investigative lead reports and not positive identifications, those working in the SNAP Unit say its facial recognition system never produces a false positive.

Another use for facial recognition searches involves unidentified persons. The SNAP Unit can help identify deceased persons who have been found but lack identification. In instances such as these, facial recognition technology can assist law enforcement officials with identifying victims of violent crimes and bring closure to families whose loved one is deceased. It can also help in solving missing persons cases. Once a potential identification has been made, additional testing (such as DNA) can help conclusively determine the person’s identity.

Two examples of the facial recognition database being used to help identify deceased persons include an incident in Muskegon County in which a body was found in a river; the person’s fingerprints were not on file and the family had not yet filed a missing person report. The SNAP system was able to provide a lead resulting in a proper identification. Similarly, the fingerprints of a drowning victim in Bay County had degraded by the time his body was located. Again, the SNAP Unit was able to produce a lead resulting in a positive identification.


Prosecutors have an independent responsibility to carefully vet evidence and information submitted by law enforcement when deciding whether to authorize charges. This is one of the reasons a warrant should never be authorized when the only source of identification for a suspect is a facial recognition lead report. Police agencies can and should use this information as a potential lead — similar to using a Silent Observer or Crimestoppers tip — but a facial recognition lead report is insufficient on its own to support charging someone.

The MSP is not the only agency with access to the various facial recognition programs. Leads from other agencies must also be closely scrutinized to ensure the same stringent protocols are followed.


PAAM has developed a best practices recommendation for facial recognition technology to help prosecutors fulfill their obligation to seek justice, protect the constitutional rights of all people, and hold offenders accountable. It recommends that prosecutors know at the warrant authorization stage whether facial recognition technology was used, consider what additional investigation was done to confirm the lead, and carefully evaluate whether witnesses or other evidence identify the suspect independent of the facial recognition lead. Examples of independent information include a close friend or relative identifying the suspect in the picture, an admission/ confession by the suspect that they were at the scene at the time of the incident, or DNA or fingerprint comparisons.

Because facial recognition technology is only a tool for lead generation, it is doubtful anyone from the MSP SNAP Unit or another agency would testify at a trial. In fact, with SNAP affirming its lead generation report does not provide probable cause (let alone proof beyond a reasonable doubt), it is doubtful that its members would be allowed to testify in most cases. That limitation requires prosecutors to ensure that other evidence of identification exists outside of and independent from leads generated by facial recognition.

PAAM’s best practice recommends a thorough investigation detailing the use of a lineup or other identification procedures including the basis for each witness’s identification of the suspect, factors that led to the identification and selection of the suspect, and additional measures taken to verify that facial recognition identification matched the physical features of the suspect. Many offices, mine included, have made this PAAM recommendation a specific part of their policy and procedure manual: A warrant shall not be authorized solely on the use of facial recognition technology that reportedly identifies a defendant. This technology is a useful tool, and can be used as part of an investigation, but the technology is not perfect and other evidence must be present in addition to facial recognition identification before a warrant should be issued.

Taking it another step further, the attorney handling the case at the preliminary examination and pretrial stages should again review the identification procedures and verify the credibility and sufficiency of the evidence. This includes interviewing witnesses to confirm whether an in-court identification can be made and, if so, whether that identification is sufficiently based on reliable evidence and procedures. As in any case, witnesses should be instructed to honestly state whether they can identify the suspect. If the person in court is not believed to be the person that committed the crime, witnesses must be advised to state that fact.

In the event that a live lineup is requested, the attorney handling the case should agree to one — provided a delay would not prejudice the case. If a live lineup would be prudent, the attorney should ensure that witnesses are not able to see the defendant in the courtroom. Lastly, facial recognition technology must not be used as the only evidence to identify the suspect.


Technology brings with it an obligation to ensure its ethical use. Michigan prosecutors take their role as ministers of justice seriously; we have a duty to protect the rights of victims justly and see that the accused are treated fairly. During my two decades as a prosecutor, the use of technology in the criminal justice system has increased tremendously from the use of in-car video to mobile phones to body cameras and now to facial recognition technology.

Every technological advance impacts how we practice and the evidence juries expect to see at trial. Whenever new technology comes forward, prosecutors are dedicated to only utilizing practices that are imperative to ensure justice in an ethical manner. At the forefront of a prosecutor’s ethical use of technology is its role in determining whether to charge an individual with a crime. Careful review and understanding the limitations of new and emerging technologies is crucial to ensuring its proper use in the criminal justice system.


“Best Practices” is a regular column of the Michigan Bar Journal, edited by George Strander for the Michigan Bar Journal Committee. To contribute an article, contact Mr. Strander at