Perth's AI face-scanning trial: what the accuracy debate misses for frontline security staff

Picture this: a patron is approached by two officers directly outside your venue doors on a Friday evening. No visible altercation, no raised voices — just a quiet pull-aside. Your door staff watch, unsure whether to step in, step back, or say anything at all. The approach was triggered by a facial recognition match. The patron is later released without charge.

That scenario is no longer hypothetical in Perth.

ABC News reports that WA Police ran a week-long live facial recognition pilot in June 2026 across Perth and Fremantle, using NEC's Neoface m40 system mounted in a police van. In seven days, more than 130,000 faces were scanned, 33 alerts were generated, and 18 arrests followed. WA Police confirmed one false positive and said a fuller data release is being coordinated. The watchlist held roughly 4,000 names spanning serious offence suspects, missing persons, and people flagged as a risk to themselves.

The public debate has focused almost entirely on whether the technology is accurate enough to trust. That is a fair debate to have. But it is not the question most relevant to a security guard standing at a venue entrance in Northbridge on a Saturday night.

The gap between policy and the person on the door

Senior operators and venue managers receive briefings. Door staff, crowd controllers, and roving guards often do not — at least not at the level of detail that would help them respond confidently when police activity overlaps with their shift.

Facial recognition policing changes the texture of street-level interactions in ways that are hard to anticipate without preparation. A standard police approach that staff recognise — responding to a reported disturbance, checking ID after a complaint — has a readable shape. A surveillance-triggered approach does not. Officers may converge on someone calmly, without obvious cause, in the middle of a venue entrance queue.

Staff who haven't been briefed on what that looks like can misread it. They might intervene unnecessarily, or fail to manage the crowd that forms around the interaction, or miss the moment when a distressed patron needs support after officers leave.

None of that is the fault of the worker. It is the result of a briefing gap.

What "one false positive" actually means at ground level

WA Police reported one confirmed false positive across 130,000 scans. Taken alone, that sounds like a well-functioning system. But the figure deserves some unpacking before operators use it to calibrate their expectations.

WA Police controlled both the timing and the framing of that number. No independent auditor has reviewed the raw trial data, and the UK's own facial recognition trials between 2018 and 2020 showed significant divergence between internally reported accuracy and independently assessed outcomes. Research from MIT's Media Lab has also documented that commercial face-scanning systems misidentify darker-skinned women at rates tens of percentage points higher than lighter-skinned men — a pattern that has direct relevance to Fremantle, where Aboriginal advocates have already questioned the geographic focus of the trial.

For a security worker, the practical takeaway is not a percentage. It is this: false positives will occur, and the person wrongly flagged will still be standing somewhere near your venue when they find out. They may be angry, frightened, or simply confused. That person becomes your problem to manage — not because you caused the situation, but because you are the one present.

Training for the moment, not just the policy

Good incident response training covers the scenarios staff are most likely to face. In venues near high foot-traffic areas in Perth and Fremantle, that list now needs to include police-initiated patron contact driven by surveillance systems.

That doesn't mean training staff to challenge or obstruct police. It means giving them a clear framework for a few specific questions: Do you maintain crowd flow around an active police interaction, or do you hold the queue? If a patron returns to the venue visibly distressed after being approached and cleared, who do you call? If an interaction happens inside your premises rather than outside, at what point does your manager need to know?

These are not complicated questions, but they require a decision in advance. Leaving them to be worked out in the moment produces inconsistent outcomes and creates records that are harder to defend if something goes wrong later.

XGuard builds this kind of operational-layer thinking into how teams are prepared for events and venue shifts — not just the procedures on paper, but the in-the-moment judgment calls that procedures don't always reach.

A simple step before your next shift

WA Police have indicated further trial data is coming, and other Australian forces are watching the outcome closely. Live facial recognition in major Australian cities is moving from pilot to probability.

Security professionals who build familiarity now — with how surveillance-triggered police approaches look, how to document adjacent incidents, and how to support affected patrons — will be better positioned than those who encounter it for the first time mid-shift.

Pro tip: Before your next event or roster in central Perth or Fremantle, ask your supervisor whether there is a local police liaison contact for the area. A brief conversation before the shift — even five minutes — can give your team a clearer picture of what activity to expect and how to respond if police and patrons intersect near your post.

The technology ran for a week in June. It will run again. The question worth asking now is whether your team knows what to do when it does.

Need protection where you are? XGuard connects you with licensed, vetted security operators in minutes — for events, residences, retail, executive protection, and fire watch. Available globally.

Source: au-abc-news — 2026-08-10