1,380,882. That is the exact number of faces scanned by police surveillance vans in Western Australia over the last five months. It’s a cold, hard statistic that stops most people in their tracks. The trial concluded on Saturday, and the vans are now parked in depots rather than idling in parking lots.
The resulting data is stark: 122 arrests and 9 false alerts. It’s a number that won’t sit comfortably with civil liberties groups across the globe.
The Perth Royal Show was the biggest single deployment site. Officers there captured 223,749 faces. It’s the most crowded public event in the state, a chaotic mix of livestock, funfair rides, and families. The police said 15 alerts were triggered during those three days of operations. Seven of those were interactions with registered sex offenders.
The Vans, The Numbers, And The Fine Print
Commissioner Col Blanch stood before reporters on Sunday to explain the mechanics of what just happened. He didn’t mince words about the scale of the operation. “I want to be very clear on those false alerts,” he said, leaning slightly into the microphone. “No one was arrested. No one made a complaint.” The tone was calm, measured, and distinctly bureaucratic.
The error rate, as calculated by the force, sits at 0.00065%. It’s a figure that looks impressive on a slide deck but meaningless in a court of law. Blanch acknowledged the technology is “not perfect” and that the force knew this going into the trial. He framed the issue as a matter of community engagement, suggesting that if a member of the public identified a suspect, police would act similarly.
Why The 0.00065% Figure Is Debatable Civil liberties advocates are pushing back hard, and they’re not wrong to be skeptical. Bill Rowlings, the chief executive of Civil Liberties Australia, called the police framing “severely gilding the lily.” He argued that the true error rate won’t be known until the 122 arrested individuals have their day in court. It’s a point that cuts to the heart of algorithmic accountability.
“If half the 122 arrested are not convicted, the ‘incorrect’ rate will be 50%,” Rowlings said in a statement. It’s a terrifying thought that forces a re-evaluation of what “accuracy” means in policing. The police scanned these faces with no reasonable suspicion whatsoever, a point that legal experts are likely to scrutinize in the coming weeks.
India’s Controversial Relevance For Indian readers, this story lands with a particular sting, and it’s not just about the technology. India is currently debating the use of facial recognition in its own law enforcement framework, with the Centre for Development of Telemetry pushing for a national database. The WA trial provides a live case study of what happens when 1.4 million people are monitored without individual suspicion. It’s a mirror that New Delhi might not want to look into.
Indian privacy advocates have long argued that mass surveillance violates the right to privacy recognized by the Supreme Court in K.S. Puttaswamy. The Indian angle here is that the global push toward “smart policing” is accelerating faster than the legal frameworks protecting citizens. If a system can flag a sex offender in a crowd of thousands in Perth, the same infrastructure could theoretically be deployed in the chaotic crowds of Mumbai or Delhi. That’s a harder sell when you consider the scale of Indian demographics.
The Algorithmic Bias Problem There is a darker undercurrent to this trial that police are keen to downplay. Facial recognition algorithms have historically shown higher false-match rates for people with darker skin tones. The WA trial did not release detailed demographic data on the 1,380,882 faces scanned, which is a massive gap. Without that data, we’re flying blind on the issue of racial bias.
Historical data from the US and UK suggests that error rates jump significantly for non-white users. If the system is biased, who gets the “interaction”? The police claim they engaged with each person treated as a match, but did that engagement feel like policing or profiling? We don’t know, and that opacity is worrying for anyone who has ever been stopped by law enforcement for no clear reason.
The Privacy Watchdog’s Silence The Office of the Information Commissioner WA, the state’s independent privacy watchdog, issued a warning before the trial began. They cited “collective privacy harms” and made it clear they were not involved in the implementation. It’s a significant line in the sand that protects the regulator from liability.
When contacted for comment after the trial, the OIC WA has yet to provide a detailed assessment. This silence is telling, as they are the only body tasked with overseeing public sector privacy. Meanwhile, the police are conducting their own internal evaluation, which promises a “review of safeguards.” It’s a classic conflict of interest: the police grading their own homework.
Frequently Asked Questions How many people were scanned in the WA trial?The Western Australia Police reported scanning 1,380,882 faces over the five-month period. The largest single deployment was at the Perth Royal Show, where over 223,000 faces were captured. What was the actual error rate?Police calculated the error rate at 0.00065% based on 9 false alerts. However, civil rights groups argue the true error rate depends on conviction rates, which are not yet available.
Did anyone challenge the arrests in court?There have been no public reports of legal challenges to the arrests at this stage. Legal experts suggest that challenges are likely to arise as the 122 cases progress through the judicial system. There have been no public reports of legal challenges to the arrests at this stage. Legal experts suggest that challenges are likely to arise as the 122 cases progress through the judicial system.
Is facial recognition legal in Australia?Australia does not have a federal law explicitly banning or regulating facial recognition, leaving it to state privacy laws. This regulatory vacuum allows trials like the one in Western Australia to proceed. Australia does not have a federal law explicitly banning or regulating facial recognition, leaving it to state privacy laws. This regulatory vacuum allows trials like the one in Western Australia to proceed.
What happens to the data after the scan?Police stated that images of people not on the alert list were pixelated in real time and deleted immediately. Data is not stored for those who were not flagged by the system. Police stated that images of people not on the alert list were pixelated in real time and deleted immediately. Data is not stored for those who were not flagged by the system.
Source: Guardian World Source: Guardian World
Source: Guardian World
