Skip to main content
Incident intelligence/SS-IR-030CASE FILE OPEN
Symbolic editorial illustration for SS-IR-030SERVANTSTACK // INCIDENT INTELLIGENCEFORENSIC IMAGE // VERIFIED FRAME
SS-IR-030 // INCIDENT REPORTAlleged

Detroit Police

Facial Recognition Match Wrongly Jails an Eight-Months-Pregnant Woman for Carjacking

EXECUTIVE BRIEF

On February 2, 2023, Detroit police arrested Porcha Woodruff, 32 and eight months pregnant, at her home as she was getting her children ready for school, charging her with robbery and carjacking.

SURVEILLANCE & POLICING
FAILURE CHAINTRACE COMPLETE
  1. 01TRIGGEROn February 2, 2023, Detroit police arrested Porcha Woodruff, 32 and eight months pregnant, at her home as she was…
  2. 02MACHINE ACTIONDecision system
  3. 03MISSING GATENamed SME review and decision audit trail
  4. 04IMPACTRights & due process
01 // INCIDENT SUMMARY

The short version

On February 2, 2023, Detroit police arrested Porcha Woodruff, 32 and eight months pregnant, at her home as she was getting her children ready for school, charging her with robbery and carjacking.

02 // KEY FACTS

Case telemetry

INCIDENT
SS-IR-030
DATE
February 2, 2023
SYSTEM
Detroit Police
LOCATION / SCOPE
Detroit, Michigan, United States
EVIDENCE
Alleged
AI ROLE
Decision system
HARM
Rights & due process
SOURCES
3 cited records
03ENTRY POINT // WHAT HAPPENED

The event

On February 2, 2023, Detroit police arrested Porcha Woodruff, 32 and eight months pregnant, at her home as she was getting her children ready for school, charging her with robbery and carjacking. The case began when investigators ran gas-station surveillance video of an unidentified woman through Detroit's facial recognition system (the city relied on DataWorks Plus) and got a hit on Woodruff. Detectives built a six-person photo lineup using a roughly eight-year-old image of her, and the victim picked her out. Woodruff was held in custody for about eleven hours, during which she had contractions and was treated for dehydration, then released on a $100,000 bond. On March 6, 2023, prosecutors dismissed the case for lack of evidence, conceding the arrest rested in part on a false facial recognition match. According to the ACLU, Woodruff was the first woman and at least the sixth person known to be wrongfully arrested in the United States because of this technology; all six were Black. It was also the third such wrongful-arrest allegation against the Detroit Police Department in three years.

04CAUSAL TRACE // AI'S ACTUAL ROLE

What the machine did

The facial recognition algorithm produced an investigative lead that was treated as if it were probable cause. The software returned a candidate match against a mugshot database and a stale photo, and rather than a qualified human treating that output as an unverified tip requiring corroboration, officers fed the algorithm's pick straight into a photo lineup and then a warrant. There was no human SME validation gate confirming that an eight-months-pregnant suspect matched a carjacking captured weeks earlier, no check that the comparison image was current, and no review of the documented racial-bias and reliability limits of the system before the match drove an arrest. The algorithm's probabilistic guess flowed end to end into a felony charge with no qualified human standing between the model output and the loss of a person's liberty.

Decision systemAutomation was a causal participant—not a decorative label for the system around it.
05BLAST RADIUS // CONSEQUENCES

Where the failure landed

An eight-months-pregnant woman was jailed for roughly eleven hours, experienced contractions and dehydration in custody, was charged with two felonies, and had to post a $100,000 bond before the case collapsed a month later. She became the sixth publicly documented person, and the first woman, falsely arrested in the U.S. via facial recognition, and the third tied to Detroit PD, intensifying national scrutiny and ACLU demands that the department stop using the technology. Woodruff sued the city of Detroit; a civil-rights claim against the warrant officer was later dismissed in 2025. Detroit subsequently changed its facial recognition policies, barring arrests based on a facial recognition match alone.

06 // EVIDENCE STATUS

Alleged

Claims reported in litigation or public allegations; not presented here as a final finding.

SOURCE RECORD UPDATED 2026-07-09

07 // SOURCE LEDGER

3 cited records

  1. 01
  2. 02
  3. 03
08CONTROL FAILURE // MISSING GOVERNANCE

Named SME review and decision audit trail

The failure pattern in this case: Automated judgment without accountable review.

09INTERVENTION POINT // HUMAN IN THE MIDDLE

The moment the path could change

A qualified reviewer tests the basis, context, and disparate impact before the decision reaches a person.

AI PROPOSES→HUMAN OWNS THE DECISION→SYSTEM EXECUTES
10CONTROL DEPLOYMENT // AUTHORITYGATE

SME routing · decision audit trail

AuthorityGate's Operational Resilience framework forbids an algorithm's output from advancing to a consequential action without a named, qualified human SME validating it at a defined gate. A facial recognition hit would be classified as an unverified investigative lead, not evidence, and the framework's change-validation gate would block it from entering a photo lineup or arrest warrant until a human reviewer confirmed independent corroboration, verified that the comparison image was current rather than eight years old, and documented that obvious disqualifiers (here, an eight-months-pregnant suspect for a physical carjacking) had been reconciled. The gate makes the human accountable for the decision, not the model: no probabilistic match crosses the threshold from "lead" to "probable cause" without a signed-off SME review on the record. That single checkpoint, the one missing here, is what stops a software guess from becoming an arrest.

RELEVANT KEYSTONE CONTROLHuman-in-the-Loop ValidationHow high-risk actions route to a named subject-matter expert who owns the go or no-go decision.
12 // THE ALTERNATIVE

Autonomy is a design choice.

See the operating model that keeps AI useful while preserving human authority at consequential moments.

Compare AgenticAI and AugmentedAI →