The first 300 seconds of an attack: lessons from the 2026 Crowded Places Conference 

One of the most impactful things said at this year’s New Zealand Crowded Places Conference in Auckland wasn’t about technology. It was a question posed to all of the security professionals in the audience: what are you going to do in the first 300 seconds of an attack at your site? 

Not what your plan says. Not what your guarding contractor is engaged to do – but what your actual on-site people will do in the first five minutes while working with an incomplete picture and a radio channel already overloaded, and before anyone else arrives to help. 

Most organisations cannot definitively answer that question. The gap between good intent, SOPs, and executable response is where most protective security risk actually sits in New Zealand. It is also where the most cost-effective improvements are available, and very few of them involve buying any new security technology.

The threat picture has been restated, not escalated 

The NZSIS has set the national terrorism threat level at Possible, an assessed 25–50 percent likelihood of a terrorist attack in New Zealand. Importantly, this is not an increase. It is a relabelling of what was previously described as “Low”, made because that word gave operators the impression there wasn’t much risk to manage. 

There is. And the profile of it matters more than the number: 

• Lone actors, often radicalised online, taking deliberate steps to avoid detection. 

• Mixed, unclear or unstable ideologies, with individuals switching between belief systems and the drive toward violence as the constant. 

• Readily accessible weapons and methods rather than sophisticated capability. 

• Limited or no intelligence forewarning. Agencies have been explicit that communities and businesses are more likely to see early indicators than security services are. 

The practical translation for anyone operating a venue, precinct, transport hub or campus is uncomfortable but simple: you should not plan on being warned. Sophisticated capability is not required to carry out a deadly attack, only malicious intent, whatever its origin: radicalisation, mental health crisis, grievance or state-sponsored influence.

Two consequences follow. 

First, the first few minutes decide the outcome. Fatalities in that attack occurred inside two minutes. No response model that assumes external help arrives in time will survive contact with reality, which is why security officers and frontline staff, not police, are realistically the first line. 

Second, hindsight is absolutely not a defence. Guidance exists. Sector frameworks exist. Self-assessment tools exist and are free. An organisation that has not engaged with them is in a materially weaker position than one that has, regardless of whether an incident ever occurs. The findings make plain that “we could not have known” does not survive contact with a coronial process when the guidance was sitting on a public website the whole time. 

The direction of travel in New Zealand 

NZ Police confirmed that the Crowded Places Strategy is under review, with a working group including Police, DPMC and NZSIS. Over the next twelve months the sector should expect: 

• A significant upgrade to the self-assessment and audit tools, a 10–12 month workstream. 

• Adoption of ANZCDC guidance with New Zealand-specific narrative, covering active offender, hostile vehicle, IED and bladed weapons scenarios, with drone guidance to follow. 

• Continued push to make Escape, Hide, Tell as instinctive as fire and earthquake drills. 

Councils are already making self-assessment mandatory as a condition of event permits. That is the shape of things: not legislation yet, but a steadily rising expectation that operators can evidence what they have done. 

Design does the heavy lifting before technology does 

One of the clearest themes of the conference was that CPTED (Crime Prevention Through Environmental Design) is a force multiplier for everything else. CPTED is passive and proactive; traditional security is active and reactive. Where the environment is designed well, cameras, guards and access control work as intended. Where it isn’t, that hardware is forced to work twice as hard for a worse result. 

Applied to crowded places, the Ministry of Justice’s seven qualities of safer spaces deliver very specific operational benefits: 

• Surveillance and sightlines. Hostile reconnaissance is harder to conduct unobserved, and abnormal behaviour stands out against expected patterns. 

• Access management and layout. People evacuate instinctively rather than by reading signage, which is what actually happens under stress. 

• Territorial reinforcement. Clear public/private transitions make unauthorised presence obvious. 

• Activity support. Genuine “eyes on the street” from ordinary daily use. 

• Physical protection. Bollards, HVM and surveillance, deployed where design has already narrowed the problem. 

The objective is not a fortress. It is a place people want to use every day that also performs when it needs to. Good protective security is largely invisible. 

Left of bang: the best 300 seconds are the ones you never have to run 

Day two shifted the frame from response to prevention. Lloyd Richardson of AccessX built his session around Left of Bang, the Van Horne and Riley text with the central proposition that it is better to detect hostile intent early than to respond to violent action later. Simple to state, hard to operationalise. 

Using Eden Park as the worked example, he stepped through the layered zones a patron moves through on approach, and where machine assistance genuinely buys back time: 

• On approach, using cameras you already own. Deep learning weapons detection can run as an overlay on existing CCTV, including relatively low-resolution streams, to identify a brandished knife or firearm before it reaches the crowd. Detection goes to trained human verifiers, and a confirmed threat can be escalated to law enforcement in seconds, with the offender tracked across the camera network rather than manually chased from view to view. The alternative is the one we all recognise from the news: excellent footage, reviewed afterwards.

• At the line, on concealed weapons. AI layered over millimetre-wave or electromagnetic sensing distinguishes a blade from a phone, a belt buckle or a cluster of keys, which is what makes high-throughput screening viable. Queues at the entrance are not a neutral inconvenience. They are an unscreened crowd standing in the open, and the argument for fast full screening is as much about dissolving that queue as it is about what gets caught. 

• On vehicles. Brief but pointed: hostile vehicle mitigation has to be certified. A planter box and a lump of concrete are not a rated solution, and the market has plenty of product that looks the part. 

Three cautions from that session are worth carrying into any procurement. 

First, false positives destroy systems. He cited an overseas airport deployment where an underpowered algorithm alerted on holstered police sidearms, sports equipment, and everyday objects until the control room was saturated and the system was simply switched off inside a week. An alert nobody trusts is worse than no alert, because it consumes the operator attention you were trying to protect and results in operators dismissing all alerts, even the genuine ones. 

Second, lift the bonnet on the algorithm. “AI” on a datasheet covers everything from genuine deep learning to simple pattern matching dressed up by marketing. Ask what is actually running, ask for detection and false-positive performance data, and ask about grey zones. A body scanner that cannot see an ankle is not a screening solution. 

Third, the human stays in the loop. Every credible deployment described used the machine to compress time and the human to make the decision. Notably, the platforms discussed deliberately avoid facial recognition and keep processing local, which is the right posture for New Zealand privacy expectations and worth writing into your requirements rather than hoping for. 

The connection between the two days at the Crowded Places conference is the important part. Day one said the first 300 seconds decide the outcome. Day two said the most valuable thing technology can do is start that clock earlier, or stop it starting at all. 

What we take from this at Nextro 

Four things we think are worth acting on, in order of return and priority. 

1. Test the first 300 seconds, realistically. Not a walkthrough. An exercise with degraded communications, an incomplete picture, an absent decision-maker, and frontline staff rather than executives making the calls. Most organisations discover their gap is authority and protocol, not equipment. 

2. Make your existing estate work harder before you buy more. Camera coverage designed for retail loss prevention or general situational awareness is rarely positioned for incident response, evacuation monitoring or post-incident evidential support. A coverage and design review against a crowded places scenario frequently produces more capability than a new capital programme would, and it is also the precondition for any analytics overlay being worth deploying. Detection algorithms cannot see what the camera was never pointed at. 

3. Treat analytics as an operational commitment, not a licence purchase. The question is never whether the demo works. It is who receives the alert, how it is verified, what authority that person holds, how it reaches Police, and what happens at 2am on a public holiday. We would rather design that workflow into your VMS and monitoring model (using Genetec Mission Control) up front than retrofit it after the first false alarm erodes confidence and a genuine incident takes place. 

4. Look outward. The strongest performers described at the conference, Newmarket Security (the security arm of Newmarket Business Association) among them, succeed because of intelligence sharing across organisational boundaries, quarterly forums, and drills run with neighbouring entities rather than in isolation. Threat information travels faster than any single organisation’s own controls. The same applies to camera networks: the most useful detection can often be on infrastructure owned by the council, the transport agency, or your neighbour. 

And a challenge worth repeating, put to the room by the Paul Patel of the Community Advisory Group: what if every organisation in the security sector helped just one community group become more resilient this year? A one-hour workshop, an emergency contacts review, a tabletop exercise. Cheap to give. Disproportionate in value. 

“Relationships create trust. Trust encourages communication. Communication improves preparedness. Preparedness builds resilience.” – Paul Patel  

Trust cannot be built during an emergency. It has to already be there. 

Nextro designs, deploys and manages unified protective security for New Zealand’s critical infrastructure and crowded places sites: video surveillance, access control, perimeter intrusion detection, hostile vehicle mitigation, AI-enabled detection, and enabling 24/7 monitoring. If you’d like a conversation about how your current environment performs against a crowded places scenario, get in touch.