A Supermarket's Face-Matching System Called a Maori Woman a Thief. Three IDs Didn't Change It.

What happened

In April 2024, Te Ani Solomon walked into a Foodstuffs supermarket in New Zealand and was told by staff to leave. The store's facial recognition system had flagged her as a trespassed shoplifter. She was not one. She offered three forms of photo identification. Staff still insisted she go.

Solomon said she caught a glimpse of the image the staff had been looking at on a phone. It appeared to show a different Maori woman wearing a cap. Solomon is Maori. That was, it seems, enough for the match to hold, at least in the moment. She described the experience as humiliating and said she felt racially discriminated against. It happened on her birthday.

Foodstuffs was running a six-month trial of facial recognition across 25 of its stores at the time, framing the program as a shoplifting deterrent. Shoppers entering those stores had no practical ability to opt out of having their faces scanned and compared against a database of trespass flags. Critics called the arrangement highly intrusive, noting that customers were giving up their biometric data whether they consented or not. The system surfaced a match for Solomon, and store staff acted on it without independent verification.

The company's public response attributed the incident to "genuine human error," a framing that places responsibility on the employee who enforced the flag rather than on the system that generated it. That explanation does not account for why three forms of government-issued photo identification were insufficient to override a machine-generated accusation in real time. Facial recognition systems have documented accuracy disparities across racial groups, with darker-skinned and Indigenous faces producing higher false-positive rates than others. Deploying such a system in a consumer retail setting, without a visible correction pathway, treats the algorithm's confidence as more reliable than the person standing in front of the camera.

The deeper problem is that no independent record existed of what the system flagged, how it generated the match, or how that confidence score compared against Solomon's actual face. She had no access to the image staff were using as a reference, no mechanism to contest the technical output while still in the store, and no clear path to a review afterward. What accountability infrastructure is meant to supply in cases like this is a provable record of what a system did, on what basis it acted, and a means for the affected person to understand and challenge it. Without that record, any store running a similar trial is one false match away from repeating the same harm.

Reported impact

Affected parties

Not publicly disclosed

Harm type

Not publicly disclosed

Scale

Not publicly disclosed

Financial impact

Not publicly disclosed

Regulatory action

Not publicly disclosed

Classification

Organization

Not publicly disclosed

AI system

Not publicly disclosed

Industry

Not publicly disclosed

Country

Not publicly disclosed

Provider

Not publicly disclosed

Incident type

Not publicly disclosed

Relevant governance controls

Governance control mapping is not available for this record.

• No controls mappedNot publicly disclosed

Control mapping is analytical. It does not state that any control would have prevented the incident.

Sources and evidence

This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.