Detroit Police Chief Admits Facial Recognition Rarely Provides an Accurate Match


DETROIT, Mich. (July 22, 2020) – During a public meeting last month, Detroit Police Chief James Craig admitted that the city’s facial recognition technology rarely provides a direct match and misidentifies people the vast majority of the time.

Detroit paid $1 million for the facial recognition software developed by DataWorks Plus back in 2017. Craig admitted that the system is appallingly inaccurate.

“If we would use the software only [to identify subjects], we would not solve the case 95-97 percent of the time. That’s if we relied totally on the software, which would be against our current policy … If we were just to use the technology by itself, to identify someone, I would say 96 percent of the time it would misidentify.”

This dovetails with other evidence revealing the inaccuracy of facial recognition technology, especially when it comes to identifying non-whites. During a test run by the ACLU of Northern California, facial recognition misidentified 26 members of the California legislature as people in a database of arrest photos.

DataWorks GM Todd Pastorini compared the system to automated fingerprint ID systems that pull up dozens or even hundreds of potential matches. He told Motherboard the system “doesn’t bring back a single candidate.”

“It’s hundreds. They are weighted just like a fingerprint system based on the probe [and what’s in the database].”

As Motherboard pointed out, this means cops are ultimately making the decision to question and investigate people based on what the software returns and a detective’s judgment.

“This means that people who may have had nothing to do with a crime are ultimately questioned and investigated by police. In Detroit, this means, almost exclusively, Black people.”

According to data released by the Detroit Police Department, officers used the facial recognition system 70 times through the first six months of the year. Of the 70 photos fed into the system, 68 were of black people. The race of the individuals in the other two photos were classified as “unidentified”

Of the 70 photos, 31 were pulled from social media accounts and 18 were captured by security cameras.

Detroit’s facial recognition system works with the vast network of cameras known as Project Green Light. The surveillance network utilizes thousands of government and private cameras throughout the city. The cameras are installed at schools, parks, apartment buildings, immigration centers, gas stations, churches, hotels, fast-food restaurants, and even in places such as addiction treatment centers and abortion clinics.

The program was implemented in 2016 and was generally popular due to the promise that it would deter and help solve crime. As the New York Times pointed out, the system is anything but covert. A flashing green light marks the location of every camera linked into a network that feeds directly into the Detroit Police Department’s downtown headquarters.

Detroit’s facial recognition system came under fire last year when Mayor Mike Duggan implied that the Detroit Police Department wasn’t using the technology in order to muddy the waters as information about the controversial program became public. As it turns out, he was using clever wordplay. After all, why would the city spend $1 million for a system it had no intention of using?

In fact, Duggan never claimed the police department wasn’t using facial recognition at all. He just said it wasn’t using it on “live stream video.”

In other words, police aren’t running facial recognition in real-time. But they are using the technology on still images plucked from reams of footage collected by cameras all around the city. As Urban Institute’s Justice Policy Center senior policy analyst Daniel Lawrence told the Detriot Free Press, this is a difference without any real distinction.

“In all my experience with facial recognition, the way the process and programming works is that it takes a still image from the video. I’m not knowledgeable of any facial recognition software that’s taking real video. It’s taking a still from a video.”

The issues with facial recognition technology go far beyond their inaccuracy. Even if the technology improves, it still poses a significant privacy threat.


recent report revealed that the federal government has turned state drivers’ license photos into a giant facial recognition database, putting virtually every driver in America in a perpetual electronic police lineup. The revelations generated widespread outrage, but this story isn’t new. The federal government has been developing a massive, nationwide facial recognition system for years.

The FBI rolled out a nationwide facial-recognition program in the fall of 2014, with the goal of building a giant biometric database with pictures provided by the states and corporate friends.

In 2016, the Center on Privacy and Technology at Georgetown Law released “The Perpetual Lineup,” a massive report on law enforcement use of facial recognition technology in the U.S. You can read the complete report at The organization conducted a year-long investigation and collected more than 15,000 pages of documents through more than 100 public records requests. The report paints a disturbing picture of intense cooperation between the federal government, and state and local law enforcement to develop a massive facial recognition database.

“Face recognition is a powerful technology that requires strict oversight. But those controls, by and large, don’t exist today,” report co-author Clare Garvie said. “With only a few exceptions, there are no laws governing police use of the technology, no standards ensuring its accuracy, and no systems checking for bias. It’s a wild west.”

With facial recognition technology, police and other government officials have the capability to track individuals in real-time. These systems allow law enforcement agents to use video cameras and continually scan everybody who walks by. According to the report, several major police departments have expressed an interest in this type of real-time tracking. Documents revealed agencies in at least five major cities, including Los Angeles, either claimed to run real-time face recognition off of street cameras, bought technology with the capability, or expressed written interest in buying it.

In all likelihood, the federal government heavily involves itself in helping state and local agencies obtain this technology. The feds provide grant money to local law enforcement agencies for a vast array of surveillance gear, including ALPRs, stingray devices and drones. The federal government essentially encourages and funds a giant nationwide surveillance net and then taps into the information via fusion centers and the Information Sharing Environment (ISE).

Fusion centers were sold as a tool to combat terrorism, but that is not how they are being used. The ACLU pointed to a bipartisan congressional report to demonstrate the true nature of government fusion centers: “They haven’t contributed anything meaningful to counterterrorism efforts. Instead, they have largely served as police surveillance and information sharing nodes for law enforcement efforts targeting the frequent subjects of police attention: Black and brown people, immigrants, dissidents, and the poor.”

Fusion centers operate within the broader ISE. According to its website, the ISE “provides analysts, operators, and investigators with information needed to enhance national security. These analysts, operators, and investigators…have mission needs to collaborate and share information with each other and with private sector partners and our foreign allies.” In other words, ISE serves as a conduit for the sharing of information gathered without a warrant. Known ISE partners include the Office of Director of National Intelligence which oversees 17 federal agencies and organizations, including the NSA. ISE utilizes these partnerships to collect and share data on the millions of unwitting people they track.

Reports that the Berkeley Police Department in cooperation with a federal fusion center deployed cameras equipped to surveil a “free speech” rally and Antifa counterprotests provided the first solid link between the federal government and local authorities in facial recognition surveillance.