Western Australia police have concluded a five-month trial of real-time facial recognition after scanning 1,380,882 faces during 100 deployments in 43 regional and metropolitan locations. The force said the trial, which ended on Saturday, led to 122 arrests and nine incorrect alerts. The deployments also led officers to engage with 146 registered sex offenders and to locate 10 people for welfare checks.
Three operations at the Perth Royal Show accounted for 223,749 scans, producing 15 alerts, three arrests and seven interactions with registered sex offenders, according to police.
A police spokesperson said the nine incorrect alerts against 1.38 million scans represented an error rate of 0.00065%. WA police commissioner Col Blanch told a Sunday media conference that officers spoke with each person involved in a false alert. He said no one was arrested and no complaint was made, and described the number as very low but not zero. Blanch said the technology was not perfect and that authorities knew that before the trial, adding that officers followed up alerts as they would public suspicions.
Civil Liberties Australia chief executive Bill Rowlings criticised the force's presentation of the error rate. He said police were overstating the system's accuracy and argued the incorrect rate should be assessed by how many of the 122 arrests lead to convictions. Rowlings said the true incorrect rate would only be known once those arrested had their day in court, suggesting that if half were not convicted the incorrect rate would be 50%. He also said the scans were carried out without reasonable suspicion.
When the trial was announced, the Office of the Information Commissioner WA, the state's independent privacy watchdog, said it was not involved in implementing the trial and warned about collective privacy harms. The office was contacted for comment.
Under trial guidelines, cameras mounted on or near marked police vehicles scanned faces in designated public areas and matched them against an alert list drawn from police databases. Police said images of people not on the list were pixelated in real time and deleted immediately without being stored. The trial had previously drawn criticism over privacy and algorithmic bias, with experts saying facial recognition can disproportionately affect marginalised groups who are already overrepresented on police watchlists, and that algorithms have historically had higher false-match rates for people with darker skin tones.
The surveillance vans have been removed from public deployment while police conduct an internal evaluation of how the technology performed and its community impacts, including a review of safeguards and whether all requirements were met.
Sources: The Guardian World.