WA facial recognition trial records 122 arrests after more than 1.38 million face scans

A five-month trial of real-time facial recognition in Western Australia ended on Saturday after police scanned 1,380,882 faces, with alerts leading to 122 arrests. The technology was deployed 100 times across 43 metropolitan and regional locations, according to WA police.

Police said alerts also prompted contact with 146 registered sex offenders and helped officers find 10 people for welfare checks. Three deployments at the Perth Royal Show accounted for 223,749 face scans. Those operations produced 15 alerts, leading to three arrests and seven interactions with registered sex offenders.

The system produced nine incorrect alerts during the trial. A police spokesperson put the error rate at 0.00065%, calculated against the total number of scans. Speaking at a media conference on Sunday, WA police commissioner Col Blanch said officers approached and questioned each person involved in a false match, but none was arrested and no complaints were made.

Blanch acknowledged that the technology was imperfect and said police had understood that before starting the trial. He compared officers’ responses to alerts with the checks they would make if someone appeared to be wanted or a member of the public reported a suspicion. He said the trial sought to identify people presenting a risk while leaving most members of the public unaffected.

Civil Liberties Australia chief executive Bill Rowlings disputed the way police presented the incorrect alert rate, accusing the force of “severely gilding the lily”. He argued that the outcome of court proceedings for the 122 people arrested should determine the assessment. In a statement, he said that if half were not convicted, the incorrect rate would be 50%. He also criticised police for conducting almost 1.4 million scans without reasonable suspicion.

Under the trial’s guidelines, cameras attached to or placed near marked police vehicles scanned faces in designated public spaces and compared them with an alert list assembled from police databases. Police said images of people absent from that list were pixelated as they were captured, then immediately deleted without storage. When the trial was announced, the state’s independent privacy watchdog, the Office of the Information Commissioner WA, said it had no role in implementation and warned of “collective privacy harms”.

Privacy and algorithmic bias concerns had already drawn criticism of the trial. Experts warned that facial recognition could disproportionately affect marginalised communities already overrepresented on police watchlists, noting historically higher false-match rates for darker-skinned people. Police have withdrawn the surveillance vans from public deployment while an internal evaluation examines operational performance, community impacts, safeguards and compliance with the trial’s requirements.