← The impacts · AItopiaOrAImageddon?

entry · p2026-09-06-260 · validated

Automated licence plate reader optical plate recognition (multi-vendor)

the city of San Francisco paid $495,000 for her seizure at gunpoint, constitutional injury, and severe emotional distress
tech
alpr-plate-recognition
passes the definition as
the reader's character recognition is a vision model fit to plate images; the hot-list comparison it feeds is a hand-written rule, and only the misreads are scored here
from
2009-03
region
US
domain
safety
side
loss
people
1 driver held at gunpoint after a plate misread (2009-03)
effect
20 minutes held handcuffed at gunpoint (2009-03) — Denise Green was ordered out of her car at gunpoint, made to kneel and handcuffed for roughly 20 minutes because the reader read a 3 as a 7; San Francisco later paid $495,000
reversible
no
who holds the switch
owner:police departments operating ALPRs
attribution
named
source kind
secondary
the statistic
EFF's roundup of ALPR-error detentions names three misread cases: Green (San Francisco, March 2009, 3 read as 7, $495,000), Jason Burkleo (Atherton, California, H read as M) and Jaclynn Gonzales with her 12-year-old sister (Espanola, New Mexico, 2 read as 7), all held at gunpoint. One study cited there found ALPRs misread the state of 1 in 10 plates. The Aurora and Contra Costa detentions in the same piece came from the hot list, not the reader.
source
https://www.eff.org/deeplinks/2024/11/human-toll-alpr-errors
proposed by
researcher (opus), brief 2026-09-06, us-surveillance
received
2026-09-06T14:27:54-04:00
harvest report
impacts/harvest/2026-09-06-us-surveillance.md

Rulings, in order

  1. 2026-09-06T14:27:55-04:00 · validatorpass — quote found on the fetched page (exact match; HTTP 200 text/html; charset=utf-8)
  2. 2026-09-06T14:28:26-04:00 · sessionseverity — Held at gunpoint, kneeling and handcuffed, because the reader misread one digit: injury risked and no recourse in the moment — severe; the city's settlement is compensation, not reversal. Only the misread is the model's; the hot-list detentions in the same source are hand-written comparison and are set aside. Reach on one driver (2009).
The proposal as it arrived (JSON)
{
 "id": "p2026-09-06-260",
 "received": "2026-09-06T14:27:54-04:00",
 "proposer": "researcher (opus), brief 2026-09-06, us-surveillance",
 "kind": "entry",
 "sha": "aefc064e3f20ddab",
 "data": {
  "kind": "entry",
  "tech": "alpr-plate-recognition",
  "tech_name": "Automated licence plate reader optical plate recognition (multi-vendor)",
  "definition": "the reader's character recognition is a vision model fit to plate images; the hot-list comparison it feeds is a hand-written rule, and only the misreads are scored here",
  "region": "US",
  "domain": "safety",
  "side": "loss",
  "from": "2009-03",
  "people": {
   "n": 1,
   "unit": "driver held at gunpoint after a plate misread",
   "date": "2009-03"
  },
  "effect": {
   "n": 20,
   "unit": "minutes held handcuffed at gunpoint",
   "date": "2009-03",
   "what": "Denise Green was ordered out of her car at gunpoint, made to kneel and handcuffed for roughly 20 minutes because the reader read a 3 as a 7; San Francisco later paid $495,000"
  },
  "reversible": false,
  "switch": "owner:police departments operating ALPRs",
  "statistic": "EFF's roundup of ALPR-error detentions names three misread cases: Green (San Francisco, March 2009, 3 read as 7, $495,000), Jason Burkleo (Atherton, California, H read as M) and Jaclynn Gonzales with her 12-year-old sister (Espanola, New Mexico, 2 read as 7), all held at gunpoint. One study cited there found ALPRs misread the state of 1 in 10 plates. The Aurora and Contra Costa detentions in the same piece came from the hot list, not the reader.",
  "source": "https://www.eff.org/deeplinks/2024/11/human-toll-alpr-errors",
  "quote": "the city of San Francisco paid $495,000 for her seizure at gunpoint, constitutional injury, and severe emotional distress",
  "attribution": "named",
  "source_kind": "secondary"
 },
 "from_file": "impacts/harvest/2026-09-06-us-surveillance.md"
}