Mayor Duggan: I oppose use of Facial Recognition Technology for surveillance

Dear Residents of the City of Detroit:

There’s been a lot of discussion and confusion the last couple weeks on the issue of the Detroit Police Department’s use of facial recognition technology. Our residents too often suffer the pain and loss of violent crime.  We expect the police to be vigorous in reducing that violence, but I wanted you to hear directly from me how I believe we need to balance that with the privacy rights of our community.

I strongly oppose the use of facial recognition technology for surveillance.

The Detroit Police Department does not and will not use facial recognition technology to track or follow people in the City of Detroit.  Period.  Detroiters should not ever have to worry that the camera they see at a gas station or a street corner is trying to find them or track them. 

DPD is not permitted to use facial recognition software for surveillance and I will never support them doing so.  The technology is just not reliable in identifying people from moving images and research has shown it is even less reliable in identifying people of color.

I have spoken to several members of the Detroit Police Commission and have encouraged them to continue this practice by formally adopting a “no surveillance” policy for facial recognition technology and providing for serious discipline for any DPD employee who violates this policy.

There have been a number of misleading reports that have confused Green Light or traffic cameras with facial recognition technology.  They are not correct.

The Green Light cameras do not have any facial recognition technology – they are standard security cameras.

The traffic cameras we are proposing to purchase do not have any facial recognition technology – they are standard traffic cameras.

I fully support the use of cameras to address the violence in this community.  I do not support the use of those cameras to conduct facial recognition surveillance.

 

Green Light cameras have been enormously successful without facial recognition

Nearly 600 businesses have voluntarily installed Green Light cameras in the last three years.

Carjackings in Detroit have dropped dramatically since Green Light started in 2016:

Total carjackings committed in Detroit

A 58% reduction in the number of carjacking victims is real progress.  Detroiters constantly heard warnings not to stop for gas in the city at night.  Today, you can see large numbers of customers at Green Light gas stations late every evening.  They aren’t perfect, but Green Light cameras have created zones of safety as prospective carjackers have learned that they are almost certain to be arrested and convicted when their crime is recorded on a high definition camera. 

 

We Expect Traffic Cameras to be Successful in Reducing Drive-by Shootings without Facial Recognition

We have proposed cameras at major intersections to fill in the holes of the Green Light system to help identify vehicles used in shootings.  This has nothing to do with facial recognition – that technology would be useless in identifying people in moving vehicles.

In January, a 3 year old boy was shot and killed on the Southfield Freeway on his way to Sesame Street Live.  It was a case of random road rage normally very difficult for police to solve.  But in this case, a gas station camera happened to catch a video of the silver Mercedes involved in the shooting as it exited the freeway and it was that video that led to the arrest of the shooter.

We continue to have far too many victims of drive-by shootings and far too many remain unsolved when police cannot conclusively identify the car.  In the first week of July, we had 12 drive by shootings with 16 victims.  4 of them died.  In each case we had a general description of the car from witnesses.  For example:

Thursday, July 4th 9:55 PM:  61 Year Old Woman was shot on Orleans near Seven Mile by a shooter in a blue Chevy Malibu.

Friday, July 5th 1:46 AM:    25 Year Old Man was shot at Seven Mile and Conant by a shooter in a Burgundy Chevy Trailblazer

As of this morning, the police have not made arrests in either case.  Had we had traffic cameras at major intersections, it is highly likely we would have license plate numbers and identifying characteristics of the Blue Malibu and Burgundy Trailblazer. 

The gun violence in this city is completely unacceptable.   Cameras at traffic intersections will help identify the vehicles of drive-by shooters and ultimately reduce the shootings.  We do not need and will not use facial recognition surveillance on the traffic cameras to accomplish this.

 

The appropriate use of Facial Recognition Technology:  learning the identity of dangerous offenders

The Detroit Police Department’s purchase of facial recognition software was approved by Detroit City Council by a 6-0 vote in 2017.  It has been used for the last two years by DPD detectives to identify dangerous offenders, without any negative incidents.  No one has proposed expanding its use. Here’s how it is used today:

Oftentimes police get a picture of an offender while committing a crime.  It could come from a citizen’s doorbell camera, a Green Light camera, or a private security camera.  Often we ask for the public’s help to identify the offender, showing their picture on the TV news.

Homicide detectives at times tried to identify an offender from a picture by spending hours looking through mug books.  In the last two years, DPD has taken the picture of that unknown offender and used the facial recognition technology to try to find matching pictures in mug books and other records.

On November 25, at 5:49 AM at a (non-Green Light) gas station on Van Dyke, a 34 year old man was shot.    The gas station’s security camera got a clear picture of the shooter, but no one was able to identify him.  A week later, the facial recognition software found a tentative match in a police mug book.  Detectives pursued that lead and quickly found that the same suspect had posted a picture of himself on a public social media page wearing the identical distinctive jacket he had worn during the shooting.  Subsequent police investigation positively confirmed his identity.

If your loved one was shot and there is a picture of the shooter, wouldn’t you expect the police to use every tool they can to identify that offender?  Police never make an arrest just because there is a facial recognition match.  But it is an important source of leads detectives can use to find the identity of the offender.  I fully support the technology’s use for that limited purpose.     

 

Summary

The Detroit Police Department has not and will not use facial recognition technology for surveillance.  No one is watching you on any camera in this city with facial recognition software.  I will not support the software ever being used in that way.

If you have committed a dangerous crime and the police have a picture of you, only then can police detectives use facial recognition software on that picture to try to determine your identity. 

The most painful moments I experience as Mayor are conversations with the families of victims who just want to know when the police are going to make an arrest in the shooting.  Those conversations are even more painful when the family knows the police have a picture of the offender and still can’t make an ID.  Facial recognition software can be very important in bringing peace to those families.

I hope the Board of Police Commissioners will adopt a policy that recognizes where this technology is helpful, but which also strictly prevents facial recognition surveillance and provides strong punishment for any abuse of that policy.  It’s my hope we can find common ground on this issue.

 

Michael E. Duggan
Mayor of Detroit