Western Australia Police scanned 1,380,882 faces during a five-month live facial-recognition trial. The agency reports 292 alerts, nine incorrect alerts and 122 arrests during its deployments. Those are striking numbers. They do not all answer the same question. 1
How many of the arrests followed directly from a facial-recognition alert? The police summary lists arrests alongside alerts but does not provide that breakdown. The question matters because an arrest that happened during an operation is not automatically an arrest caused by the technology.
Three numbers, three meanings
Police say the cameras were used 100 times in 43 locations between June and October 2026. The technology compared faces in public spaces with an alert list, and officers assessed a match before deciding whether to engage someone. The agency says an alert alone is not grounds for an arrest. 1
A scan is an attempted comparison. An alert is a result the system flags for review. An arrest is a later action by an officer. One does not necessarily produce the next, and the summary's totals do not show every link in the chain.
Police also reported nine incorrect alerts and expressed that number as 0.00065% of scans. That describes identified incorrect alerts divided by all scans. It is not the same measure as the share of alerts found incorrect, and neither figure alone tells us how many matches the system missed, how independently each result was reviewed or how often deployments affected different groups. 12
What would make the result clearer?
A useful evaluation would show how many alerts were reviewed, how many led officers to approach someone, how many were confirmed or found incorrect, and what happened after those encounters. It would also distinguish arrests made because of an alert from arrests made for other reasons during the same deployment.
Police say the trial will undergo an internal evaluation examining performance, operational effectiveness, community impact and governance. Their published figures are a starting point for that evaluation, not a substitute for the missing connections. 1
The difference is important for both sides of the debate. A system may help locate someone, and a false alert may also subject an innocent person to scrutiny. To judge either outcome fairly, the public needs to see how a camera result became—or did not become—an action.
References
- Live Facial Recognition Technology. (n.d.). https://www.wa.gov.au/organisation/western-australia-police-force/live-facial-recognition-technology
- Police use facial recognition to scan one million faces in WA trial leading to 122 arrests. (n.d.). https://www.theguardian.com/australia-news/2026/nov/02/wa-police-facial-recognition-technology-trial