Latest / Case Cracked / Episode 8- The Facial Recognition Flaw That Freed the Wrongly Accused
Transcript
- Ronan Wolfe: Welcome to Case Cracked: The Clues That Led to Justice. I’m Ronan Wolfe, your AI detective, here to guide you through stories where overlooked clues, new technology, and sharp thinking converge to expose the truth. Today’s episode tackles a case that’s both groundbreaking and unsettling—one where an algorithm pointed fingers, and a man almost paid with his future. Ronan Wolfe: In early 2023, Detroit police made headlines when they arrested 34-year-old Jordan Myles outside his workplace. He was stunned, handcuffed in front of coworkers, and accused of armed robbery. The case against him? A surveillance photo and a facial recognition match linking him to a gas station holdup days earlier. Ronan Wolfe: The photo, captured by a low-resolution security camera, showed a man in a black hoodie brandishing a pistol. The image was grainy, pixelated, and lacked clear distinguishing features. Yet the city’s facial recognition software flagged Jordan Myles as a likely match. The system, which cross-referenced DMV photos with the blurry image, returned a confidence score of 87 percent. Ronan Wolfe: That was enough for detectives to move forward. Despite Myles having no prior criminal record, a solid alibi, and no physical evidence connecting him to the scene, he was booked and held for 11 days before a judge granted bail. His life was already unraveling—he lost his job, endured public shaming, and faced an uncertain future. Ronan Wolfe: But one piece of overlooked data would soon change everything. Myles’ attorney, Lianne Carter, began digging into the facial recognition report. She noticed something odd—the match didn’t just suggest Myles; it also flagged two other individuals with similar facial geometry, both with different skin tones and facial hair. The algorithm had returned multiple results, and investigators had chosen to pursue the one they felt looked "most likely." Ronan Wolfe: Carter filed a motion to obtain the full match report. When it was released, independent AI ethics researchers were brought in to examine the algorithm’s process. Their findings were eye-opening. The software had a known error margin of 12 percent with Black male subjects—something the department had been warned about months earlier. Ronan Wolfe: But this wasn’t just about bias. It was about procedure. Further investigation revealed that the original surveillance footage had been slightly distorted due to lens warping from the gas station camera. The distortion caused facial metrics like jaw width and eye spacing to skew—enough to fool the algorithm. Ronan Wolfe: Then came the break. A separate gas station camera, located diagonally across the street, captured a different angle of the suspect fleeing the scene. This footage had been overlooked because it belonged to a privately-owned station not part of the city’s surveillance network. Ronan Wolfe: When reviewed, this footage clearly showed that the suspect had a distinctive tattoo on his neck—something Jordan Myles didn’t have. Facial reconstruction specialists confirmed the man in the secondary footage bore only a superficial resemblance to Myles. Ronan Wolfe: Armed with this new evidence, the case against Myles was dropped. A public apology followed, but the damage was done. He had been publicly branded a criminal based on an AI recommendation and little else. Ronan Wolfe: The fallout didn’t stop there. The case ignited national debate over the use of facial recognition in policing. The Detroit Police Department suspended its use of the software pending a full audit. Meanwhile, Myles’ legal team launched a civil lawsuit, alleging negligence and violation of constitutional rights. Ronan Wolfe: During trial proceedings, an even deeper layer emerged. The facial recognition company had issued internal guidance warning that confidence scores below 95 percent should be used only as leads, not as grounds for arrest. The department had ignored those warnings, relying instead on visual comparisons by investigators. Ronan Wolfe: Myles was awarded a $2.7 million settlement. More importantly, his case prompted legislative changes in Michigan, mandating stricter oversight and transparency for any law enforcement agency using biometric identification tools. Ronan Wolfe: This case isn’t just a cautionary tale—it’s a milestone. It shows that technology can assist justice, but cannot replace due process. Algorithms are tools, not arbiters of truth. And behind every scan is a life, a name, a story that deserves more than a percentage match. Ronan Wolfe: Thanks for joining me on Case Cracked: The Clues That Led to Justice. Subscribe and follow for more stories where digital evidence, critical thinking, and a relentless search for truth bring justice into focus. I’m Ronan Wolfe, and I’ll see you next time—where another case waits to be uncovered, and another truth demands to be heard.