The Privacy Commissioner says facial recognition technology in supermarkets has potential safety benefits, despite raising significant privacy concerns.
An inquiry from the Office of the Privacy Commissioner into facial recognition trialled by Foodstuffs in the North Island found any business using facial recognition tech, or considering doing so, must ensure the system is set up correctly to stay within the law.
Commissioner Michael Webster says the system raised privacy concerns, like the unnecessary or unfair collection of a customer’s information, misidentification, technical bias and its ability to be used for surveillance.
The commissioner found the live technology model used in the trial was compliant with the Privacy Act.
“These issues become particularly critical when people need to access essential services such as supermarkets. FRT [facial recognition technology] will only be acceptable if the use is necessary and the privacy risks are successfully managed,” Webster said.
The Foodstuffs trial ended last September and ran in 25 supermarkets.
About 226 million faces were scanned during the trial, including multiple scans of the same person, and 99.999% of those were deleted within one minute.
The trial raised 1742 alerts, 1208 were confirmed matches to store watchlists – databases made from images of people of interest to a store.
In December 2024, a woman took her case to the Human Rights Review Tribunal after she was wrongly kicked out of a Rotorua supermarket, claiming the technology was discriminatory.
There were nine instances of someone being approached by staff, but misidentified as the wrong person during the trial. In two cases, the shopper was asked to leave.
All nine instances were attributable to human error, and were outweighed by the benefits of using facial recognition, justifying its use.
The inquiry found while the level of intrusion to customers’ privacy was high because every visitor’s face was collected, the safeguards used in the trial reduced the intrusion to an acceptable level.
Webster said there was still work needed to improve the safety and efficiency of facial recognition software for New Zealand, as it had been developed overseas and not trained on a local population.
He said the commission could not be completely confident the technology had addressed issues on technical bias, and that it had the potential to negatively impact Māori and Pacific people.
“This means the technology must only be used with the right processes in place, including human checks that an alert is accurate before acting on it.
“I also expect that Foodstuffs North Island will put in place monitoring and review to allow it to evaluate the impact of skin tone on identification accuracy and store response, and to provide confidence to the regulator and customers that key privacy safeguards remain in place,” Webster said.
The safeguards included immediately deleting images that did not match with a store’s watchlist, setting up the system to only identify those whose behaviour was seriously harmful, like violent offending, not allowing staff to add images of people under 18 or those thought to be vulnerable to the watchlist and not sharing watchlist information between stores.
Match alerts were verified by two trained staff members to make sure a human decision was part of the process, the inquiry report said, and access to the facial recognition system and its information was restricted to authorised staff.
Images collected were not permitted to be used for training data purposes, the report said.
General counsel for Foodstuffs North Island Julian Benefield said the goal behind the FRT trial was to understand whether it could reduce harm while respecting people’s privacy, saying it had succeeded in doing so.
“Retail crime remains a serious and complex problem across New Zealand,” he said.
“Our people continue to be assaulted, threatened and verbally abused, and we’re committed to doing all we can to create safer retail environments.”
Benefield said privacy was at the heart of the trial.
He said an independent evaluator found the trial prevented more than 100 cases of serious harm, including assaults.
“We have worked closely with the Office of the Privacy Commissioner and listened to their feedback.
“We welcome the OPC’s feedback on areas for improvement and will carefully consider their recommendations, including the need to monitor accuracy, before we make any decisions about future permanent use.”
Retail NZ chief executive Carolyn Young said the trial showed the technology had made a measurable impact in reducing harm and improving safety in stores.
She said retailers across the country had been watching the trial with interest and a number were investigating FRT for their own operations in the near future.
“Retailers are crying out for proactive solutions that prevent crime and enhance the safety of their staff and customers. Our members continue to face high rates of violence and crime, putting both their employees and the public at risk, as well as threatening the financial sustainability of retail businesses.”
- RNZ