External Control Layer

Police arrest a man on a facial recognition lead

A facial recognition result labeled “not probable cause to arrest” led to an arrest after, according to the man’s lawsuit, a lineup built from the match was shown to someone who had not witnessed the theft.

Date
January 2020
Kind of system
Predictive or matching model
Outcome
Harm

What happened

In 2019 a still from a store’s surveillance video of a watch theft was run through facial recognition, which returned Robert Williams’s expired driver’s license photo as a lead. According to his lawsuit, detectives put his photo in a lineup shown to a loss-prevention contractor who had not been in the store and had been told there was a match. The lawsuit also alleges that the warrant request did not give the prosecutor or the magistrate the video. He was arrested at home on 9 January 2020 and held for about 30 hours.

The prosecutor later apologized, saying the case “should not have been issued based on the DPD investigation,” and the police chief called it sloppy work. In the settlement of 28 June 2024, which admits no liability, Detroit agreed that an arrest cannot be based on a facial recognition lead and a lineup and “must be supported by additional independent reliable evidence.” It also agreed to audit arrests and arrest warrants that followed a facial recognition lead, from February 2017 through August 2023.

What allowed the action to take effect

  • G6 No independent evidence

    The action, or the approval of it, was based on the system’s own output or on a single unverified input. Evidence that would have shown the error was never gathered, was thrown away, or was available and not consulted.

    The complaint alleges that the only evidence beyond the match was a lineup identification by someone who “was not an eyewitness” and had been told of the match. The settlement now requires that an arrest “must be supported by additional independent reliable evidence.”

    Kind of evidence: allegation. Sources: U.S. District Court, Eastern District of Michigan, 23 June 2023; U.S. District Court, Eastern District of Michigan, 28 June 2024. References: Complaint ¶¶ 93, 96, 99; Directive 307.5-5.3.

    Related guidance: P6 Signer independence and P4 Reviewable policy. Contract clause C-6.

  • G7 Review in name only

    A person was supposed to catch the error but lacked the information, time, training, or practical means to do it, so the review did not work as a check.

    The complaint alleges the lead report left out the match’s confidence score and the other candidates, and that neither the prosecutor nor the magistrate who approved the warrant was given the video.

    Kind of evidence: allegation. Source: U.S. District Court, Eastern District of Michigan, 23 June 2023, ¶¶ 81, 118, 121.

    Related guidance: P5 Human authority. Contract clause C-5.

  • G9 Warning with no authority

    Signs that the action was wrong existed before or during the harm, but none of them could stop or hold the action, and nothing required anyone to act on them.

    The lead report stated: “THIS DOCUMENT IS NOT A POSITIVE IDENTIFICATION. IT IS AN INVESTIGATIVE LEAD ONLY AND IS NOT PROBABLE CAUSE TO ARREST.” The warning was not enforced in the steps that followed.

    Kind of evidence: allegation. Source: U.S. District Court, Eastern District of Michigan, 23 June 2023, ¶ 80, quoting the report.

    Related guidance: P3 Fail-closed and P8 Sequence awareness. Contract clauses C-3 and C-8.

Considered and left out

G8 No human decision
Detectives, a prosecutor, and a magistrate made the decisions in this case, and the gaps marked here concern what those decisions were based on.

Entries with gaps in common

Sources

Cite this entry

External Control Layer. (2026). Entry 8: Police arrest a man on a facial recognition lead. In Control Gap Record (Version 0.3). G-14, Inc. https://externalcontrollayer.ai/record-8

@misc{ecl_cgr_8,
  author = {{External Control Layer}},
  title = {Entry 8: Police arrest a man on a facial recognition lead},
  howpublished = {Control Gap Record, version 0.3},
  publisher = {G-14, Inc.},
  year = {2026},
  url = {https://externalcontrollayer.ai/record-8},
  note = {Released 2 November 2026}
}

Uses Control Gap Classification 0.2.

Correct this entry

Email ecl@g14.ai with “ECL record 8” in the subject line.

Write the email with the subject ECL record 8