Foodstuffs releases privacy assessment for FRT trial in NZ grocery stores

New Zealand grocery retailer Foodstuffs has published a Privacy Impact Assessment for its trial of in-store facial recognition technology.

The facial recognition trial, which began in late October, has FRT deployed in three PakNSave supermarkets in Christchurch run by Foodstuffs South Island. It aims to explore FRT as a way to reduce crime and incidents of abusive or threatening behavior toward Foodstuffs employees, including assault, the use of weapons, incidents of verbal abuse and incidents of disorderly conduct.

In its proportionality assessment, FSSI says a similar trial run by Foodstuffs North Island in 2024 was effective in “reducing serious threatening behaviour of repeat offenders in a comparable retail setting.” That trial scanned more than 225 million faces and raised 1742 alerts, with 1200 confirmed against watchlists. In its assessment, the Office of the Privacy Commissioner said that “while the level of privacy intrusion was high because every visitor’s face is collected, the privacy safeguards in the trial reduced it to an acceptable level.”

Auckland’s Auror gets a chance to showcase FRT tech

On the technical side, the report says, the FSSI facial recognition system is “made up of the Imagus (v10) FRT software from Vix Vizion Pty Limited,” an Australian firm, “which will be hosted and managed by FSSI as a centralised system in its own secure instance of AWS cloud,” as well as the Auror platform from Auckland-based Auror Limited, “which will be hosted by Auror Limited in Microsoft Azure and which integrates and interoperates with the FRT Software via an application programming interface (API) made available by Vix Vizion as part of the FRT Software.”

Until recently, Auror focused on vehicle licence plate recognition, but has expanded its toolkit to include facial recognition, saying the technology’s accuracy has improved dramatically.

FSSI aims to operate in compliance with the principles laid out in New Zealand’s Privacy Act 2020. In terms of collecting personal information, its system uses biometric cameras to record facial images of every person who walks into a store, and assign a biometric template to each. “These templates are matched in real time against a centrally managed Watchlist of identified extreme and high-risk offenders,” says the report. “If there is no match, no data is sent by the FRT Software to the Auror Platform and the facial image and biometric template is deleted from the FRT Software within seconds of when the image enters the FRT Software.”

A match with the watchlist triggers an alert containing the name of the identified individual, time, store location and camera image, as well as a behavior classification: “High Risk” or “Extreme Threat.” In the North Island trial, the template had to be 92.5 percent accurate for it to be flagged.

According to RNZ, teenagers are the worst repeat offenders in Foodstuffs stores, and the biggest threat to FSSI staff, with six out of 10 of the top offenders across its retail network identified as being under 18. That said, minors are not included on the facial recognition watchlist, so cannot trigger alerts.

Article Topics

criminal ID  |  data privacy  |  facial recognition  |  New Zealand  |  New Zealand Privacy Commissioner  |  retail biometrics