西澳大利亚警察从6月22日起在珀斯和曼德拉试用实时面部识别技术，利用标志性警车对行人面部进行扫描，并与4000人监视名单进行比对[1]。试验首周扫描超过13万张面孔，产生33次警报，导致18次逮捕[1]。
该试验采用了日本NEC公司的NeoFace Watch系统[1]。警察声称仅有一次误报，误报率为0.0008%[1]，但根据33次警报的实际数据计算，误报率实际上为3-6%[1]。英国国家物理实验室的研究发现，当降低匹配阈值以提高识别灵敏度时，深肤色人群会面临不成比例的误报风险[1]。
试验的进行引发了对隐私法治理的重大疑问[1]。该试验缺乏独立评估、透明的成功标准和外部监督[1]。西澳大利亚新颁布的《2024年隐私和负责任信息共享法》是澳大利亚首部直接规范自动化决策中个人信息使用的法律[1]。国际人权法要求对隐私的干涉仅在合法、必要和均衡时才被允许[1]。2020年英国法院曾裁定南威尔士警察的实时面部识别使用违法[1]，这为西澳试验的合法性提出了挑战。
Western Australia Police commenced a trial of live facial recognition technology on June 22, 2024, deploying marked police vehicles to scan pedestrian faces in Perth and Mandurah against a watchlist of approximately 4,000 individuals [1]. During the trial's first week, the system scanned over 130,000 faces, generating 33 alerts that resulted in 18 arrests [1]. The technology utilizes Japan-based NEC's NeoFace Watch system [1].
The trial has raised significant transparency and accuracy concerns. Police claimed only one false match occurred, suggesting a 0.0008% error rate, yet the 33 total alerts generated would indicate a false positive rate of 3 to 6 percent [1]. Research by the UK's National Physical Laboratory found that when matching thresholds are lowered, false positive rates disproportionately impact individuals with darker skin tones [1]. The trial lacks independent evaluation, clear success metrics, and external oversight [1].
The deployment occurs as Western Australia introduced the Privacy and Responsible Information Sharing Act 2024, making it Australia's first legislation directly regulating personal information use in automated decision-making [1]. The legal framework reflects international human rights standards requiring that privacy intrusions be lawful, necessary, and proportionate [1]. A 2020 British court ruling had previously found South Wales Police's live facial recognition use unlawful, establishing a precedent relevant to such programs [1].
