Police in Western Australia scanned 1,380,882 faces over five months using vans equipped with facial recognition cameras, The Guardian reported. The camera generated 292 alerts, of which 122 resulted in arrests and nine were incorrect, according to figures published by the Western Australia Police Force on November 2, 2026. The force also published a privacy impact assessment, an equality impact assessment and a district-by-district breakdown of deployments.

Similar disclosures from Delhi Police about its facial recognition vehicles would help establish how the technology is being deployed and monitored. MediaNama reported on a Delhi Police vehicle marked “Facial Recognition System” and “South District” parked outside Select Citywalk mall in Saket. MediaNama could not find any public records detailing where these vehicles are deployed, how their use is authorised or how often the systems generate incorrect matches.

What did Western Australia’s trial involve?

The trial ran from June 22 to October 31, 2026, and involved around 100 deployments across 43 locations in Perth and regional Western Australia, according to the force.

Cameras “mounted on or positioned in the vicinity of a marked police vehicle” compared faces against “a strictly controlled alert list consisting only of lawfully held police images“, and images of people not on the list “were automatically pixelated in real time and were not saved“, the force said.

The Japanese company NEC is named as the vendor in the force’s privacy impact assessment. The assessment states that, “if endorsed”, the force would use NEC’s NeoFace Watch technology and NeoFace M40 algorithm, which it says are also used by London’s Metropolitan Police. The trial “draws on UK policing models, principally from the Metropolitan Police Service“, the assessment says.

What did the trial find?

  • Scale: 1,380,882 faces scanned and 292 alerts generated.
  • Outcomes: 122 arrests, 146 registered sex offenders “identified and engaged with”, and 10 people located for welfare checks.
  • Errors: nine incorrect alerts, which the force put at 0.00065%.
  • One event: three deployments at the Perth Royal Show scanned 223,749 faces, generating 15 alerts and three arrests.
  • Alert lists: across deployments up to October 11, faces were compared against alert lists averaging about 8,000 entries, according to the district breakdown.

“It’s a very, very, very low number. It’s not zero. We’d love it to be zero,” Western Australia Police Commissioner Col Blanch said of the incorrect alerts, The Guardian reported.

No one was arrested on a false alert, although officers did approach each person, he said. While the force evaluates the trial, the vans have been withdrawn from public deployment.

How was the error rate calculated?

The 0.00065% figure divides the nine incorrect alerts by all 1.38 million faces scanned. The nine incorrect alerts accounted for approximately 3% of the 292 alerts generated by the system, or roughly one in 32. The district breakdown, which runs to October 11, also lists 34 alerts as “Unconfirmed (person not engaged)”, separate from the eight it counted as incorrect by then.

The framing was questioned by civil liberties groups. The chief executive of Civil Liberties Australia, Bill Rowlings, said that the police were “severely gilding the lily” and that “we will only know what the incorrect rate is when the 122 people arrested have their day in court”, The Guardian reported.

Sarah Schwartz, legal director at the Human Rights Law Centre, said every false match is “a real person who may be stopped, questioned or investigated because an algorithm got it wrong“.

What rules did the force publish?

According to the force’s website, each deployment must be approved by an officer of the rank of Superintendent or above. The approving officer must consider ‘whether the deployment is necessary and proportionate’, ‘why the chosen location is appropriate’ and which categories of people should be included on the alert list. The page says “an alert is not, by itself, grounds to arrest or exercise powers”, and that the public is told through signs, posters with QR codes and advance notice online.

The two assessments add more detail:

  • Opting out: “individuals cannot opt out of being captured by CCTV if they decide to walk through the OLFR zone,” the privacy impact assessment says, referring to overt live facial recognition.
  • Highest risks: the assessment rated information security as the highest privacy risk, and found a “moderate residual risk” on transparency, citing “insufficient public awareness” and “limited stakeholder and community engagement”.
  • Children: the equality impact assessment requires “enhanced authorisation requirements for persons aged under 14 years” before they are put on an alert list.
  • Religious headwear: officers are to be briefed that they have “no general legal power to require removal of religious headwear or coverings solely for passing an LFR system”.
  • Sensitive places: deployments near schools, places of worship and medical facilities get extra scrutiny.

MediaNama could not identify any Western Australian law specifically governing police use of facial recognition systems. The force’s webpage says the technology “operates consistent with existing policing powers and the legal frameworks concerning taking images of people in public places”. Schwartz told The Guardian that the force’s public policies “appear heavily borrowed from overseas models“.

What did the privacy regulator say?

Western Australia’s Office of the Information Commissioner said in a media release on July 9, 2026 that it “was not invited to participate in any formal consultation process with WA Police and did not provide input into the design of the trial”.

The privacy regulator noted that facial images constitute sensitive personal information under the state’s Privacy and Responsible Information Sharing Act 2024. It also warned of “scope creep”, saying that “a shift to the use of covert technology or fixed widespread use (for example, using existing CCTV cameras) would significantly increase privacy risks“.

What has Delhi Police disclosed?

By comparison, Delhi Police has disclosed little information about its facial recognition systems. An affidavit submitted to the Supreme Court in August is among the few publicly available documents describing their use. Over seven days at the Jantar Mantar protests, 2,873 people were flagged by the system as having prior records, MediaNama reported. MediaNama could not find a published deployment register, accuracy figures, alert-list sizes, standard operating procedures, or privacy and equality impact assessments for Delhi Police’s facial recognition vehicles. MediaNama also did not observe any signs at the Saket mall indicating that facial recognition technology was in use.

Delhi Police has not responded to the questions MediaNama sent to its South District office on November 3.

What MediaNama is asking Delhi Police

MediaNama will be filing an application under the Right to Information Act seeking:

  • The number of Delhi Police vehicles fitted with facial recognition systems, district-wise, with the date each was inducted, its vendor and cost, and the tender or contract under which it was bought.
  • The deployment log of the South District vehicle, including its deployment at Select Citywalk on November 1, 2026, and the rank of the officer who authorised each deployment.
  • Copies of any standard operating procedure, order or guidelines governing where and how these vehicles are deployed, including signage and how long images are kept.
  • Copies of any privacy impact assessment or accuracy and bias testing of the software, and the name of the software used.
  • Month-wise numbers of deployments, matches generated, matches confirmed, matches found incorrect, and arrests following a match in 2026.
  • The categories of people whose images are stored in the database against which the vehicles compare faces, and the number of records it contains.

Also read:

The post What Western Australia’s facial recognition van trial disclosed, and what Delhi Police hasn’t appeared first on MEDIANAMA.