AURORA | Aurora Police got another tentative OK Monday night from city lawmakers for permission to use facial recognition software and systems to help identify suspects and solve crimes.
“The technology would provide opportunities for enhancement of productivity, increase our ability to solve crimes,” Police Commander Chris Poppe of Aurora’s District 3 said during a study session on Monday. “At the end of the day, the community is safer.”
The request from police to city council, required by state law, was moved from a study session to the council floor with one council member objecting. The proposal was passed out of a council committee in September.
Poppe assured the city council that these are the same tools they would use for recognition, but the AI is more likely to be accurate than human recognition alone. He added that facial recognition would be strictly used in the aftermath of crimes to identify suspects, witnesses or even victims, and not for ongoing surveillance or preemptive monitoring.
“I am very concerned about the use of AI in this manner,” Councilmember Alison Coombs told Poppe at the meeting on Monday. She was the lone dissenter to move the proposal ahead. “The scope is extremely broad, the number of potential witnesses and the ability to access their data, whether or not they’re interested in having that happen, is deeply concerning to me.”
Poppe said that there is no intention to access anyone’s private data and that the police department will utilize a mugshot database and publicly available information, such as publicly accessible social media pages.
If approved, the department would be committed to a multi-layered training and oversight program, he said. Their operators will receive training from the FBI, as well as software-specific onboarding. Detectives who receive leads from facial recognition results will also receive specialized training on how to evaluate and act responsibly in response to such leads.
To address the potential for false positives, the police department would use a human operator to review the algorithm’s candidate list. They will also undergo a peer review, also known as a “meaningful human review,” as mandated by state law. Any information will also be reviewed and approved by a supervisor before being forwarded to the lead detectives.
Facial recognition results alone will never be treated as probable cause for arrest, Poppe said.
“Never would a match on a photo be the only piece of evidence that would take us to that point of probable cause that we get to when we arrest someone,” he said. “There are going to be several other investigative techniques that are used to link that person to the crime.”
Councilmember Gardner said he had concerns about wrongful identification, citing national cases where individuals were misidentified and detained based on flawed AI results. He asked what safeguards were in place to prevent similar errors in Aurora.
Pete Schulte, city attorney, said that under Colorado law, facial recognition results do not constitute reasonable suspicion or probable cause on their own. Investigators must corroborate any match with traditional investigative work, he said. Otherwise, it would be like someone mistakenly identifying a mugshot on the news and the police not verifying it is the right person before arresting them.
Using facial recognition has been controversial for several years, with some critics claiming that people of color have a higher rate of misidentification than white people. The aspect that seems to be stressed most by lawmakers and police departments is that it cannot be used as singular evidence, as was the case in a 2023 incident involving a man named Randal Quran Reid in Georgia, who was held in jail for six days after being misidentified by facial recognition technology.
Critics also have concerns about civil liberty and privacy regarding the technology, and some groups like the ACLU have asked police departments to stop using it altogether, like in Detroit, after an eight-month pregnant Black woman was held in jail for hours after being misidentified for a robbery and carjacking.
In 2022, the Colorado Legislature established guidelines for the use of facial recognition by police, requiring agencies to adopt policies, submit accountability reports and obtain approval from their governing bodies, Poppe said.
For Aurora, this means the city council must formally approve the implementation, and the public must be informed about it.
“Prior to that, we were using facial recognition, and it didn’t have a lot of guidelines,” Poppe said in a previous Public Safety, Courts and Civil Service Policy Committee meeting Sept. 11. “It was across the country, not being used necessarily responsibly. We weren’t providing accountability or much transparency. So this gave us some framework to move forward.”
The police department has already posted accountability reports on its website and opened a public comment portal. By law, police must maintain that feedback channel open for at least 90 days and hold at least three community meetings before proceeding.
Aurora Police will evaluate and incorporate that information into their final implementation, Poppe said.
“This is the start of the community input process,” Poppe said. “We’ve put a lot of effort into making sure this is done responsibly, transparently and with accountability.”
Aurora’s draft policy also outlines prohibited uses. If fully approved, police would not be allowed to use facial recognition for live surveillance, immigration enforcement, harassment or ongoing surveillance, Poppe said.
Use for marketing or commercial purposes is also barred, unless authorized by a court order. Poppe said that Nordstroms, for example, does use their cameras for facial recognition tracking for advertising and other purposes.
As required by state law, Aurora police officials are asking city leaders to authorize the use of facial recognition technology, a step that would formalize a program the department has already been building for nearly three years, according to Poppe during the previous meeting.
Poppe presented the proposal during a city council, describing it as a “deliberate and paced” effort shaped by state regulations, outside consultants and national best-practice standards.
If city lawmakers do approve it, facial recognition would be added to the department’s existing biometric tools, such as DNA and fingerprinting, Poppe said. Investigators often collect video or still images from doorbell cameras, as well as those from businesses or city cameras, if individuals and businesses are willing to provide them, Poppe said. Currently, those images are circulated through bulletins or Crime Stoppers, with the hope that someone will recognize the suspect.
Now, the police department is hoping to utilize software as an additional tool for this purpose. If approved, they will be able to take an image and compare it to a “volume of images” that they already have, like mugshots, or from the internet, including social media pictures, Poppe said.
Aurora police propose using two widespread systems, Lumen and Clearview AI, Poppe said.
Lumen is a statewide database of mugshots from Colorado jails, with which the city already partners. Clearview AI is a private company that scrapes publicly available images from social media and the internet.
Both systems would be used only after investigators establish reasonable suspicion in an ongoing case, he said. Matches generated by the software would be treated only as investigative leads, not as probable cause for arrest.
About a dozen specially trained officers would operate the software, with broader training planned for roughly 140 detectives over time.
The department estimates Clearview AI would cost about $32,000 in the first year, $36,000 the second year and $67,000 in the third. Training expenses would add to the total, Poppe said. Aurora’s participation in the Lumen database carries no additional cost.
The measure now goes to the council floor for first reading, possibly this month. The request for permission must pass a first and second reading to gain final approval.