How would you know?

In October last year, thieves broke into the Louvre in broad daylight and walked out with roughly $102 million in crown jewels. It's since become one of the most talked-about heists in living memory, and most people assume the story is about a security system that failed. It isn't, not quite. The alarm worked. It went off the moment the thieves broke in.

What failed was everything after that.

The museum's control room didn't have enough screens to keep every camera feed on display, so when the alarm triggered, there was nobody actually watching the right one. By the time a staff member manually found the correct camera and zoomed in to see what was happening, around eight minutes had passed. The thieves were gone. Camera coverage of the building had gaps, too — only a fraction of the museum's rooms had any coverage at all, and the one camera near the point of entry happened to be facing the wrong way. When guards and police did respond, they reached the actual break-in point about thirty seconds after the thieves had already left, partly because it took time to work out exactly where in a long, narrow gallery the breach was happening.

The alarm detected the intrusion, correctly and immediately. But detection that nobody's watching, feeding a response that arrives minutes too late, doesn't achieve very much. That gap — between a system correctly noticing something and someone actually being in a position to act on it — is the whole story.

What detection is actually for

It's tempting to think of detection as simply "finding out something happened." That's true, but it undersells the point. The value of detection isn't the finding out — it's the speed of finding out, paired with someone genuinely watching who can act on it. The Louvre's alarm did its job in the first second. The eight minutes that followed are what turned a detected break-in into a completed one.

Good detection isn't just a sensor that trips. It's a sensor that trips, feeding a person or a system watching in real time, capable of doing something about it before the moment has passed.

Cameras, then and now

When people think about tightening up their security, the instinct is almost always the same: put a camera up. It's understandable — cameras are visible, tangible, and they genuinely do a great job of Deter. Someone casing a site thinks twice when they see a lens pointed at them.

But historically, cameras have done very little for Detect. A camera records. It doesn't notice, and it doesn't tell anyone anything, until a person sits down and watches the footage — usually after the fact, usually because something has already gone wrong. Seeing yourself get robbed in the playback is cold comfort. You've got evidence. You didn't get a phone call while it was happening.

That's starting to change. Modern cameras with AI-based analytics can now genuinely detect — recognising a person or vehicle where there shouldn't be one, in real time, and reporting that straight through to a monitoring station, the same way a traditional sensor would. The camera that used to exist purely to capture evidence after the fact is increasingly doing double duty, picking up detection work that used to belong entirely to PIRs, beams, and reed switches. It's a genuinely useful shift — but it's an addition to good detection, not a replacement for having it at all.

"If they want to get in, they'll get in"

We hear this a lot, and honestly — it's often true. A determined intruder with the right tools and enough time can beat most physical barriers and most deterrents.

But that's the argument for detection, not against it. There is no response without detection. You cannot dispatch a guard, and you cannot call the police, about a break-in you don't know is happening. If deterrence doesn't hold, detection is what's left standing between an intrusion and somebody being in a position to do something about it.

And when a detection system reports an event, the response behind it usually isn't automatic — it's checked. A monitored control room verifies what's happening, often using camera footage tied to the alarm, before deciding what response is warranted: a guard sent to site, and where a genuine intrusion is confirmed, police. Detection is what starts that chain. Nothing happens without it — which is exactly where the Louvre's control room fell short, and where a monitored intruder detection system is built to hold up.

Where this fits

Deterrence is about not needing anything to happen at all. Detection is what matters the moment deterrence doesn't hold — the difference between an alarm nobody's watching and one that rings straight into a monitored response. It's also why, most nights, a monitored control room doesn't call anyone at all — and that silence is the system working exactly as it should.

This is the second piece in our look at the Red Flag 4D Framework — Deter, Detect, Delay, Detain — the layers that make up a genuinely resilient security approach. [Read part one: Deter →]

If you want to know how detection would actually hold up on your own site — whether an intrusion would be caught, verified, and acted on in real time, not eight minutes later