Height of failure: NYPD facial recognition wrongfully arrests man 8 inches taller than suspect

The New York Police Department wrongfully arrested Trevis Williams after facial recognition software identified him as a suspect in a public lewdness case, despite him being eight inches taller and 70 pounds heavier than the actual perpetrator. The case highlights the dangerous combination of flawed AI technology and inadequate police protocols, particularly how algorithmic bias can lead to wrongful arrests of Black individuals.

What happened: NYPD’s facial recognition system generated six potential matches from grainy CCTV footage of a February incident, all of whom were Black men with facial hair and dreadlocks.

• Investigators acknowledged the AI results alone were “not probable cause to arrest” and that anyone identified was only a potential suspect.
• Despite this caveat, detectives included Williams’ photo in a lineup, where the victim confidently identified him as the perpetrator.
• Williams was arrested on April 21, held in jail for more than two days, and was located 12 miles from where the crime occurred.

The human cost: Williams repeatedly denied involvement during questioning, telling police “That’s not me, man, I swear to God, that’s not me.”

• A detective dismissively responded: “Of course you’re going to say that wasn’t you.”
• The charges against Williams were finally dismissed in July, and the entire investigation was closed without resolution.

The bigger pattern: This incident represents at least the fourth documented case of facial recognition leading to wrongful arrests of Black individuals.

• Three similar cases occurred in Detroit, prompting legal advocates to push for stricter guidelines on facial recognition use in police lineups.
• The NYPD currently has no safeguards governing how facial recognition results should be used in criminal investigations.

Why this matters: The case exposes how police departments with massive resources—the NYPD’s budget approaches $6 billion—can deploy sophisticated surveillance technology without adequate oversight or accountability measures.

• Between 2007 and 2020, the NYPD spent over $2.8 billion on surveillance technologies, including facial recognition systems in use since 2011.
• The incident demonstrates how algorithmic bias combined with unreliable eyewitness identification can create a perfect storm for wrongful arrests, particularly affecting communities of color.