Florida Courts Could Decide How Police Use Facial Recognition Tech

The use of the technology by detectives working a drug case ended up in court and could lead to new rules around the powerful software.

  • Facebook
  • LinkedIn
  • Twitter
  • linkText
  • Email
(TNS) — The First District Court of Appeals is due to break new legal ground in determining whether police are allowed to use facial recognition software to identify suspects without ever notifying them of the technology.

The court battle, regarded by researchers as the first of its kind in the country to consider how the surveillance tool can be used in a criminal case, is being waged over a statewide biometric database of faces run out of the Pinellas County Sheriff's Office — one that extends to anyone with a Florida driver's license despite their criminal histories, or lack thereof.

Undercover detectives from the Jacksonville Sheriff's Office accessed the system through an intermediary in September 2015 to generate a lead after hitting a dead end in their search for the perpetrator of a $50 crack buy. More than two years later, state prosecutors with the Florida Attorney General's office are making novel arguments to defend the agency's use of the controversial technology.

The arguments are in response to Willie Allen Lynch's appeal of his May 2016 conviction for the sale of cocaine. The appellate judges will consider the case on numerous grounds — chief among them: whether the state was required to turn over photos of other matches returned by the face-matching software, and whether the subsequent identification process met legal standards.

The first issue centers on so-called "Brady material," named after a Supreme Court case that established the requirement that prosecutors turn over any evidence that might exonerate the defendant.

Lynch's attorneys are making the case that photos of four other potential suspects produced by the biometric database as potential matches would meet that definition. Lynch even requested those photos prior to trial after finding out they existed only by virtue of deposing detectives and a crime analyst months into the case.

Prosecutors hadn't proactively disclosed the way Lynch was identified by detectives, and his arrest report further obscured that identification, saying police used a mugshot system, which was only partially accurate. The mugshot system was linked to the facial recognition software.

The state's argument that prosecutors were not required to turn over the photographs even when asked centers on the fact that the undercover detective who identified Lynch considered only one of them — presented to him by a Jacksonville Sheriff's Office crime analyst, Celbricah Tenah.

"The trial court reasonably concluded that the evidence was not exculpatory because the undercover detectives only relied upon the single photograph sent by Tenah to identify the drug seller," prosecutors wrote in their brief.

The Lynch case was tried under former State Attorney Angela Corey. The current state attorney, Melissa Nelson, recently said her office has not set up policies directing line attorneys on how to manage facial recognition identifications in criminal cases because the issue doesn't come up often enough.

The Jacksonville Sheriff's Office also does not have a policy explicitly dealing with facial recognition. Lt. Chris Brown, of the Professional Oversight Unit, said the agency treats the technology as an investigative tool to generate leads, and its utilization falls under other policies and state law.

But privacy advocates and researchers who study the technology worry about a lack of specific safeguards to prevent misuse of the powerful technology .

In the Lynch case, the state's argument asserts that the other photos "played no role" in the detectives' identification of Lynch as the suspect.

"Moreover, there is no evidence that Tenah expressed doubt in the result of her search based on the other photographs," the brief said.

Lynch's lawyers argued that the trial court never held a hearing to determine whether the evidence was, in fact, exculpatory, despite Lynch requesting them, and that the photos would have been vital to their client's defense.

"Evidence that there were other men who fit the same facial profile as [Lynch] would have been significant in [Lynch's] attack on his identification as the seller," Lynch's attorneys wrote.

The state's failure to disclose them, they said, prejudiced Lynch's ability to defend himself against the charge and denied him due process.

"The error was not harmless and undermines confidence in the verdict," the lawyers wrote.

Remarkably, despite having no legal training, Lynch, a lifelong addict, crafted legal arguments behind bars contending that the way detectives identified him — based on a single photograph presented by a crime analyst — was overly suggestive.

Typically, double-blind lineup procedures and other safeguards are used in the identification of suspects in criminal cases to prevent witnesses from being influenced by detectives, inadvertently or otherwise.

In Lynch's appeal, state prosecutors have conceded that "a single photograph of a suspect shown to a witness is generally considered highly suggestive," but contended that the case was outweighed by "the totality of the circumstances," namely that the detectives provided Tenah with three photos of the suspect taken during the drug deal.

"The three photos of the suspect at the drug transaction, [Lynch's] presence in the courtroom, and the photograph of [Lynch] from his arrest established mitigated the risk of misidentification by the detectives," prosecutors wrote.

Lynch's lawyers said that even beyond the "highly suggestive" procedure of Tenah showing the detective only the photo of Lynch, the crime analyst also told the detectives she thought Lynch was the drug seller and provided his criminal history, which included drug sales.

"If a police officer showed a single photo of a suspect to an eyewitness, told the eyewitness that he thought his suspect was the one who committed the crime, and told the eyewitness that the suspect in the photo had a history of committing the same kinds of crimes, then this court, or any court, would be hard-pressed to find there was not a substantial likelihood of misidentification," the defense attorneys wrote.

"The process used in the instant case was the functional equivalent."

Clare Garvie, one of the country's foremost researchers in police surveillance technology, said that the state's contention that the other photos were not exculpatory is a "huge red flag," highlighting the frontier nature of the legal landscape being argued.

Further, Garvie added that the argument that the identification process was not unduly suggestive because the detectives were certain Lynch was the suspect "makes no sense."

Tenah, the crime analyst, played a central role in identifying Lynch, Garvie said, even though she was not a witness to the crime and was relying on algorithmic software she did not fully understand.

"This is just one example of the deeply concerning implications of the use of face recognition in law enforcement investigations," Garvie said.

Facial recognition software has come under fire by academic researchers such as Garvie for being mysterious in several regards. Accuracy has been called into question, as have methods and safeguards to make sure the systems work properly.

Other researchers have said the software has built in racial biases. Lynch is an African American male.

The court case has been ongoing and will stretch longer still. Prosecutors filed their response to the appeal in September, and the defense countered it two months later. The case is now awaiting a decision, which could still take several months.

©2018 The Florida Times-Union (Jacksonville, Fla.) Distributed by Tribune Content Agency, LLC.

  • Facebook
  • LinkedIn
  • Twitter
  • linkText
  • Email