WA Police Trial Live Facial Recognition in Public

ABC News reported that WA Police is using NEC's NeoFace m40 product. WA Police's official trial page says the cameras are deployed overtly near a marked police van, officers assess any generated alert before engaging a person, and non-matched data is deleted automatically and immediately.

Biometric Update reported that the five-month pilot will be evaluated before decisions on future use, and that the trial plan does not commit police to permanent cameras, covert deployment, or integration with wider CCTV systems.

The distinction between live one-to-many matching and conventional CCTV matters technically and legally. Conventional cameras may record a scene for later review, whereas live facial recognition creates biometric templates from many people in a public area and searches for potential matches against a defined database in near real time. Human confirmation can reduce the risk that an automated alert alone triggers action, but it does not remove the privacy implications of processing faces of people who are not suspects.

The Office of the Information Commissioner WA said it was aware of the trial but was not invited to a formal consultation process or asked to contribute to its design, according to Biometric Update. The watchdog cautioned against collective privacy harms.

ABC News reported criticism from privacy experts, legal advocates and Aboriginal advocates, including questions about the opacity of the watchlist and whether deployment locations could disproportionately affect First Nations people. The Guardian reported that critics have also raised concerns about false alerts, algorithmic bias and function creep, the expansion of a system beyond its initial purpose.

University of Western Australia law lecturer Anna Zenz told IDTechWire that facial recognition datasets can embed racial and gender biases and warned that errors can produce discriminatory outcomes. The retrieved sources do not provide demographic accuracy results for the specific NEC deployment in Western Australia.

For ML and data-governance practitioners, this case illustrates a recurring evaluation gap in public-sector biometric systems: aggregate outcomes such as arrests and alerts do not by themselves establish performance across demographic groups, lighting conditions, camera angles, or watchlist quality. A rigorous assessment would benefit from documented false-positive and false-negative rates, subgroup testing, retention controls, audit logs, and independent oversight before the privacy and accuracy tradeoffs can be assessed fully.

The trial is being watched by other Australian law-enforcement agencies, ABC News reported. Its eventual assessment could add evidence to an Australian policy debate already shaped by the Privacy Commissioner's finding that Bunnings' earlier facial-recognition use interfered with customer privacy.