Police AI surveillance trial draws rebuke from privacy experts

Western Australia Police have launched an Australian-first trial of live artificial intelligence facial recognition technology, drawing criticism from privacy experts and legal advocates.

The mass surveillance pilot, rolled out in Perth and Fremantle since June, uses cameras linked to facial recognition software to scan individuals and compare their faces against a police watchlist in real time. In its first week, the system scanned over 130,000 people, generating 33 alerts and leading to 18 arrests.

Police are using NEC's Neoface m40 product, comparing scanned faces against a watchlist of approximately 4,000 individuals, including those accused of serious offenses, missing persons, and individuals deemed a risk to themselves or others. WA Police have reported promising early results, noting that images of people not on the database are deleted almost immediately and appear pixelated to officers.

However, concerns have been raised regarding the opacity of the watchlist and the locations of the trial sites, with some suggesting they may disproportionately target First Nations people. Experts like Melbourne Law School associate professor Jake Goldenfein highlight the known accuracy issues of facial recognition technology, particularly for people of color and women, due to algorithms trained on predominantly white and male datasets.

While WA Police declined to release further data beyond the initial week's results, citing a coordinated release of information, the trial's expansion and its implications for public privacy are under close scrutiny by law enforcement agencies nationwide.

“

Historically, these systems have performed worse for people of colour and for women, and especially worse for women of colour.