1.38 Million Faces Scanned: What the WA Test Shows

TVS News AI, editorially supervised, presents 1.38 Million Faces Scanned: What the WA Test Shows.

1,380,882 faces were scanned during a five-month trial of real-time facial recognition in Western Australia. According to WA Police, the trial led to 122 arrests, but also resulted in nine false alarms. The outcome was therefore not just a technical success: the test also measured how reliable the system performs in real-world conditions.

For you as a member of the public, this means that cameras and surveillance vehicles can now not only capture individual faces, but compare faces with databases during an operation. The Australian trial shows how quickly this can translate into police checks—and how important it is not to downplay false alarms as positive matches.

What the Five-Month Facial Recognition Test Revealed

The test in Western Australia used surveillance vehicles that, according to the report, were deployed 100 times across 43 regional and metropolitan locations. The technology captured a total of 1,380,882 faces.

The number initially sounds like clear evidence of the technology’s usefulness. However, it primarily describes the scale of the trial. The available information does not provide details on which matches led to which arrests, how police proceeded with the results, or how many of the 122 arrests were directly attributable to a database match.

The second part of the results is therefore important for assessing the trial: nine false alarms. They show that a system cannot treat every match as a hit. The report also says that officers made contact with 146 registered sex offenders, making the potential application more concrete—but it does not establish that all those contacts were triggered exclusively by the automatic facial recognition system.

122 Arrests—but Nine False Alarms

WA Police therefore considers the trial a success: more than one million people were scanned, 122 arrests resulted, and the nine false alarms remain clearly identified as an error rate in the report.

It is precisely this combination that makes the case interesting. A technology deployed across 1.38 million faces does more than contribute to investigations. It also produces false alarms that police agencies must review. Whether such a ratio is sufficient cannot be determined from the published figures alone. The report gives no details about additional checks, manual reviews or the consequences of misidentifications.

For the debate over its use, this means that the 122 arrests point to the technology’s potential usefulness in everyday investigations. The nine false alarms argue against adopting it blindly. Both belong in the same assessment.

100 Deployments Across 43 Locations: The Regional Trial

It is particularly striking that the technology was not tested in only one central location. The 100 deployments extended across 43 regional and metropolitan locations. The system was therefore tested under different conditions rather than exclusively in a controlled environment or at a single transport hub.

This makes the findings somewhat easier to contextualize: results from a single environment are difficult to generalize. The broad distribution, meanwhile, provides a real-world impression. However, the report does not explain how police selected the locations or whether particular areas or population groups were affected differently.

According to The Guardian, the test ended on Saturday. The available information does not state what regulatory or political decision will follow. That is not because the trial lacked success: its results include hard numbers, but not a complete evaluation of how the technology performed in practice.

What the Test Means for Everyday Life

The case makes it clear that real-time facial recognition is no longer merely a technical or scientific issue. Across 43 locations and during 100 vehicle deployments, it became part of concrete police work. This can change day-to-day investigations, but it also raises questions about reliability.

The most important conclusion from the available figures is therefore: Facial recognition can produce useful matches while still generating false alarms. Both sides belong in every serious assessment. Citing only the 122 arrests hides the uncertainty. Focusing only on the nine false alarms obscures the technology’s potential investigative value.

Conclusion: A Real-World Test With Clear Figures and Open Questions

Western Australia has documented an exceptionally large trial: 1,380,882 faces, 122 arrests and nine false alarms—over five months and across 43 regional and metropolitan locations.

For the debate about its use, this is a robust starting point, but not a final answer. The figures demonstrate its reach, potential usefulness and error rate. However, they do not tell you how matches were handled in practice or what consequences misidentifications had for the people affected.

You can find more information on surveillance, security and digital responsibility in our related TVS guides.