A five-month trial of real-time facial recognition technology in Western Australia scanned more than 1.38 million faces and led to 122 arrests, WA police said, as civil liberties advocates criticised how the force framed its error rate.
The trial, which concluded on Saturday, involved 100 deployments across 43 regional and metropolitan locations. WA police said 1,380,882 faces were scanned by cameras mounted on or near marked police vehicles in designated public areas, matching them against an alert list drawn from police databases. Alongside the 122 arrests, the technology prompted officers to engage with 146 registered sex offenders and locate 10 people for a welfare check.
Deployments included three operations at the Perth Royal Show, where 223,749 faces were scanned, resulting in 15 alerts, three arrests and seven interactions with registered sex offenders.
The system generated nine incorrect alerts against the total 1.38 million scans, which WA police said amounted to an error rate of 0.00065%. At a media conference on Sunday, WA Police Commissioner Col Blanch said officers engaged with each individual. “I want to be very clear on those false alerts. No one was arrested. No one made a complaint. Each one did result in an interaction with that person and asking them relevant questions,” Blanch said. “It’s a very, very, very low number. It’s not zero. We’d love it to be zero. But I think the important thing for us is to engage with our community just like we would if a human thought someone was a wanted person.”
Blanch acknowledged the technology was “not perfect” and that authorities “knew this going into this trial”, but maintained that officers followed up on alerts in the same way they would if a member of the public raised suspicions.
However, the chief executive of Civil Liberties Australia, Bill Rowlings, said WA police were “severely gilding the lily” by claiming the incorrect alert rate was 0.00065% and suggested it should instead be calculated by comparing how many of the arrests resulted in convictions. “We will only know what the incorrect rate is when the 122 people arrested have their day in court,” he said. “If half the 122 arrested … are not convicted, the ‘incorrect’ rate will be 50%,” he said in a statement, adding that police had performed the almost 1.4 million scans “with no reasonable suspicion whatsoever”.
Police said images of people not on the alert list were pixelated in real time and deleted immediately without being stored.
The trial previously triggered backlash over concerns about privacy and algorithmic bias. Experts say the technology could disproportionately affect marginalised groups already overrepresented on police watchlists, with facial recognition algorithms historically having higher false-match rates for people with darker skin tones. When the trial was announced, the Office of the Information Commissioner WA (OIC WA) said it was not involved in implementing the trial, warning of “collective privacy harms”.
The surveillance vans have now been removed from public deployment while the force conducts an internal evaluation into how the technology performed on the ground and its community impacts, including reviewing the safeguards and whether all requirements were met.
Analysis
Why This Matters
- The trial shows the scale at which police can deploy live facial recognition in public spaces, scanning more than a million people without individual suspicion, raising questions about mass surveillance and consent.
- The dispute over the error rate highlights how statistics can be framed differently, and the real accuracy of the system for identifying wanted individuals will only be tested through court outcomes.
- The trial’s privacy and bias concerns, including higher false-match rates for darker skin tones, will inform future adoption of the technology by other police forces in Australia and abroad.
Background
Real-time facial recognition technology matches faces captured by cameras against a watchlist of wanted persons. Critics, including privacy advocates and civil liberties groups, have long warned that the technology can produce false matches that disproportionately affect marginalised communities. The WA trial is among the first large-scale deployments of its kind by an Australian police force.
Key Perspectives
WA Police Commissioner Col Blanch: said the trial “focused on using modern technology responsibly to identify people who presented a risk, while allowing the vast majority of people to go about their day without any impact”. He noted the technology was not perfect but that officers followed up on alerts as they would with a public tip-off.
Civil Liberties Australia chief executive Bill Rowlings: said the force was “severely gilding the lily” by calculating the error rate as 0.00065%, arguing the true incorrect rate would only be known after the 122 arrested people have their cases heard in court.
What to Watch
- The outcome of WA police’s internal evaluation, including whether they recommend permanent deployment or expanded use.
- Court cases arising from the 122 arrests, which will determine how many scans were indeed accurate.
- Any response from the Office of the Information Commissioner WA, which had warned of “collective privacy harms” and was not involved in the trial’s implementation.
Checked
12 claims in this story checked against 1 source: 11 supported.
Show each claim
- SupportedBlanch said, “No one was arrested. No one made a complaint,” after false alerts.Australia news | The Guardian: ““No one was arrested. No one made a complaint. Each one did result in an interaction with that person and asking them relevant questions.””
- SupportedBlanch called the number of false alerts very low, said it was not zero and said police wanted it to be zero.Australia news | The Guardian: ““It’s a very, very, very low number. It’s not zero. We’d love it to be zero.””
- SupportedRowlings said the incorrect rate would only be known when the 122 people arrested had their day in court.Australia news | The Guardian: ““We will only know what the incorrect rate is when the 122 people arrested have their day in court,” he said.”
- SupportedRowlings said that if half of those arrested were not convicted, the incorrect rate would be 50%.Australia news | The Guardian: ““If half the 122 arrested … are not convicted, the ‘incorrect’ rate will be 50%,” he said in a statement”
- SupportedRowlings said police performed the scans “with no reasonable suspicion whatsoever”.Australia news | The Guardian: “saying police had performed the almost 1.4m scans “with no reasonable suspicion whatsoever”.”
- SupportedThe headline says the WA trial scanned more than one million faces and led to 122 arrests.Australia news | The Guardian: “More than a million people have had their faces scanned in a trial of real-time facial recognition technology in Western Australia, resulting in 122 arrests”
- SupportedCivil liberties advocates question the 0.00065% error rate and say the incorrect rate will only be known when those arrested have their day in court.Australia news | The Guardian: “The chief executive of Civil Liberties Australia, Bill Rowlings, said WA police were “severely gilding the lily” by claiming the incorrect alert rate was 0.00065% and suggested it should instead be calculated by comparing how many of the arrests resulted in convictions. “We will only know what the incorrect rate is when the 122 people arrested have their day in court,” he said.”
- Partly supportedThe opening says WA police said the trial scanned more than 1.38 million faces and led to 122 arrests.Australia news | The Guardian: “WA police said under the five-month trial, which concluded on Saturday, the technology scanned 1,380,882 faces after 100 deployments across 43 regional and metropolitan locations.”
- SupportedThe five-month trial concluded on Saturday and involved 100 deployments across 43 regional and metropolitan locations.Australia news | The Guardian: “WA police said under the five-month trial, which concluded on Saturday, the technology scanned 1,380,882 faces after 100 deployments across 43 regional and metropolitan locations.”
- SupportedThe trial scanned 1,380,882 faces, and the system generated nine incorrect alerts, an error rate of 0.00065%.Australia news | The Guardian: “A police spokesperson confirmed the system generated nine incorrect alerts against the total 1.38m scans, meaning there was an error rate of 0.00065%.”
- SupportedThe trial led officers to engage with 146 registered sex offenders and locate 10 people for welfare checks.Australia news | The Guardian: “Alongside the 122 arrests, the alerts led officers to engage with 146 registered sex offenders and locate 10 people for a welfare check.”
- SupportedThree Perth Royal Show operations scanned 223,749 faces and resulted in 15 alerts, three arrests and seven interactions with registered sex offenders.Australia news | The Guardian: “The deployments included three operations at the Perth Royal Show, where 223,749 faces were scanned, resulting in 15 alerts, three arrests and seven “interactions” with registered sex offenders.”