During a five-month trial across Western Australia, police scanned nearly 1.4 million faces in public places using real-time facial recognition cameras. While authorities reported that the technology helped officers make 122 arrests and locate people for welfare checks, the system also produced nine false alerts that led officers to stop and question innocent members of the public.
The operations spanned 43 regional and metropolitan locations, using cameras mounted on or placed near marked surveillance vehicles. Faces in public crowds were scanned and matched against an alert list drawn from police databases. Police Commissioner Col Blanch acknowledged that the system was imperfect, but stressed that officers handled false alerts carefully and that no one was wrongfully detained or lodged a complaint. Authorities also noted that images of individuals not on the alert list were pixelated in real time and immediately erased.
However, civil liberties advocates and privacy watchdogs raised serious objections to the trial. Bill Rowlings, chief executive of Civil Liberties Australia, argued that scanning crowds without reasonable suspicion infringed on public freedom and contended that the real accuracy of the system will only become clear if the 122 arrests lead to convictions in court. The state's privacy commissioner also warned of collective privacy harms, while experts pointed to known risks of algorithmic bias, which can result in higher error rates for marginalized groups.
The surveillance vans have been removed from the streets while the police force evaluates the trial. Investigators are reviewing community impacts and examining whether existing safeguards were sufficient before determining whether the technology will be used again.