Concealed Weapons Detection Systems Concealed Weapons Detection Technology Compared, What a Concealed Weapon Detector Costs, and Why No Camera Is One
Two completely different products get sold under this phrase. One is a screening portal at a door that senses metal on a body. The other is AI that watches your existing cameras. Only the first can find a hidden gun, and we sell the second. This page is explicit about that, because the fastest way to waste a security budget is to buy the wrong one.
Runs on any ONVIF or RTSP camera you already own. Visible-weapon alerts plus intrusion, loitering, vehicle, and license plate detection.
- Finds concealed guns
- Screening portals only
- Cameras find
- Visible weapons only
- Vendors publishing accuracy
- Zero portal vendors
- Biggest hidden cost
- Staffing the lane
- Camera analytics
- $39 to $42 per cam
A concealed weapons detection system is a screening device at a controlled entrance that senses a weapon hidden on a person's body or in a bag, usually by reading how metal disturbs a magnetic field, and flags where on the body it sits. It is not a camera. No visible-light camera, however good its AI, can see a pistol inside a waistband or a backpack, because there is nothing for it to see. Camera analytics detect a weapon once it is drawn and in view, which is a genuinely different job at a genuinely different price.
Concealed means hidden, and hidden means invisible to a lens
The word doing all the work in this category is "concealed." A gun in a waistband under a shirt reflects no light to a camera. There is no partial outline to classify, no muzzle, no grip, no barrel. The image a camera receives is a person wearing a shirt. This is not a limitation of any particular algorithm or a problem that a larger training set will solve later. It is a property of visible light.
That is why entrance screening exists at all. A screening portal does not look at a person, it measures them. The common approach uses low-intensity electromagnetic fields and reads the distortion that ferrous and conductive mass creates as someone walks between two columns, then infers whether that signature looks like a firearm rather than a phone or a set of keys. Because it senses through fabric and nylon, it finds what a lens cannot.
Camera analytics, including ours, work on the other side of that line. Once a weapon is out and visible in frame, software can classify it and raise an alert in about a second, across every camera you own, at every hour, without a person watching a wall of monitors. That covers parking structures, loading docks, perimeters, corridors, and every door you did not put a screening lane on, which in practice is nearly all of them.
Most serious buyers end up running both, and the ones who are disappointed are almost always the ones who were sold one as a replacement for the other. If a salesperson tells you their camera AI detects concealed weapons, that single sentence tells you what the rest of the proposal is worth.
- ×That our camera analytics detect concealed weapons. They do not, and neither does any competitor's.
- ×A detection accuracy percentage for any screening portal. Not one vendor in the category publishes a verifiable one.
- ×A per-lane price for a portal. These are quoted per deployment and the public figures come from bid documents, not rate cards.
- ×That any product is "AI-powered" in a way that changes the physics of what a sensor can and cannot reach.
Concealed weapons detection technology compared
Five things get sold into the same budget line. They do not do the same job, and the differences show up on day one of operation rather than in the proposal.
| Technology | Finds concealed? | Throughput | Staffing | Coverage | Honest weakness |
|---|---|---|---|---|---|
| Free-flow screening portal | Yes | Vendors cite up to 3,600 per hour | Attendant per lane, plus resolution staff | One doorway per lane | Alarm resolution is where the labor actually goes |
| Walk-through metal detector | Yes, for metal | Slower, single file, pockets emptied | Attendant per lane | One doorway per lane | Alarms on keys, belts, phones, laptops |
| Handheld wand | Yes, for metal | Very low | One operator per person screened | Wherever a person stands | Does not scale past a trickle of visitors |
| X-ray bag scanner | Bags only | Moderate | Operator plus bag handler | One lane | Nothing on the body itself |
| Camera AI, including ours | No | Unlimited, it is passive | None added | Every camera you own | Only sees a weapon once it is drawn |
Throughput figures are vendor marketing claims, not independently verified test results. Read the next section before you rely on any of them.
Nobody selling a screening portal publishes a detection accuracy rate
We went through the public materials for every major vendor in this category. Here is what each one puts its name to, and what it leaves out. If you have seen a percentage attributed to one of these products, check whether the vendor said it or whether a reseller, an integrator blog, or a news story did.
| Vendor | Type | Publishes detection accuracy? | What it does publish |
|---|---|---|---|
| Evolv Express | Screening portal | No, and it is now restricted from doing so | Throughput. A November 2024 FTC order bars it from making accuracy, false alarm, speed, and labor cost claims. |
| Motorola Solutions CWD | Screening portal, powered by Evolv | No | Up to 3,600 people per hour, and a labor cost reduction claim in older collateral. |
| CEIA OPENGATE | Screening portal | No | A low nuisance alarm claim, 16 hours of battery, setup in under a minute, no accuracy figure. |
| ZeroEyes | Camera AI | No | Human verification in its own operations center. Full DHS SAFETY Act Designation, October 2024. |
| Omnilert | Camera AI | No | Speed only, described as fractions of a second. Full SAFETY Act Designation, March 2025. |
| Actuate | Camera AI | Yes, with a condition attached | 99% when a gun is brandished for five seconds. A separate 95% figure is false positive reduction. |
| Scylla | Camera AI | Yes, with a source | Probability of detection above 96% from US Army field testing, under 0.1 alerts per camera per day. |
| Surveillant | Camera AI | No, and we will not invent one | Published pricing at $39 to $42 per camera per month, and a free plan for one camera. |
Why the silence is the story
A detection rate only means something with the test conditions attached: which firearm, carried where on the body, at what walking speed, through what fabric, at what sensitivity setting. Change the sensitivity and you trade missed weapons against nuisance alarms directly. A single headline number hides that trade, which is a fair reason for a vendor to decline to publish one. The problem is what fills the vacuum. Buyers compare a fabricated figure for one product against a conditioned figure from another and think they are comparing like with like.
What to ask for instead
Ask for the sensitivity profile the vendor will deploy at your site and what it is tuned to miss. Ask for the nuisance alarm rate measured at a comparable site with comparable traffic, in alarms per thousand people rather than as a percentage. Ask how many secondary screenings per hour that rate implies, then staff for that number. Ask for the claims in writing, on current letterhead, because marketing collateral in this category is frequently years old. That last point is not hypothetical, as the Evolv order below shows.
The FTC order every buyer in this category should read first
On November 26, 2024 the Federal Trade Commission announced a proposed settlement order with Evolv Technologies over its weapons detection marketing. The FTC alleged Evolv made false or unsupported claims about what its scanners could detect and which harmless items they could ignore, and unsupported claims that its use of artificial intelligence made the systems more accurate, more efficient, or more cost effective than traditional metal detectors.
The order restricts Evolv from making claims in a specific and unusually broad list of categories: the ability to detect weapons, the ability to ignore harmless personal items without visitors removing them from pockets and bags, weapons detection accuracy including any comparison to traditional metal detectors, false alarm rates including any such comparison, the speed at which visitors can be screened, and claims about reducing labor costs. A limited group of K-12 school customers, described as roughly 8% of the company's customers, received the right to cancel their contracts. The settlement included no monetary relief.
Read that list again, because it is effectively the entire marketing surface of a screening portal. Throughput, no-need-to-empty-pockets, ignores-your-keys, better-than-a-metal-detector, and saves-you-guards are the five things every product in this category is sold on. A federal regulator looked at one vendor's evidence for all five and required it to stop.
This matters beyond one company for a practical reason. Evolv's technology is resold under other brands, most notably as Motorola Solutions Concealed Weapons Detection, and older channel collateral for those rebadged products still carries claims in the categories the order covers. The order binds Evolv Technologies, not every partner who printed a brochure in 2023. So the burden falls on you: get the claims that matter to your decision restated in writing, dated, from whoever is actually selling to you.
- •Ability to detect weapons
- •Ignoring harmless items without emptying pockets or bags
- •Weapons detection accuracy
- •False alarm rates
- •Screening speed
- •Reducing labor costs
Source: FTC announcement and proposed order, November 26, 2024. Nothing here is legal advice.
What a concealed weapon detector actually costs
No portal vendor publishes a rate card, so the only trustworthy public numbers come from bid documents that a public body had to disclose. Here is what those show, and where the money really goes.
$78,810
A single indoor screening lane over 48 months in the Buffalo Public Schools solicitation, about $1,642 per month. A dual outdoor configuration in the same document ran $153,290, about $3,194 per month.
4 to 5 FTE
What it takes to keep one post covered around the clock once you account for shifts, relief, leave, and turnover. Fully loaded, that is comfortably past $300,000 a year for a single 24/7 lane, and it recurs forever.
$39 to $42
Per camera per month for our platform, on cameras you already own, adding no headcount. A hundred-camera site costs less per year than one screening lane costs to staff for a single quarter.
The mistake that shows up in budget reviews a year later is treating a portal as a capital purchase. The hardware is the small half. Every screening lane needs someone attending it during every hour the door is open, plus enough additional staff to resolve alarms without a queue forming, and alarm resolution scales with visitor volume rather than with the number of weapons found. Before you compare quotes, work out the staffing hours your entrance actually needs and price those first. If that number kills the project, it was going to kill it in year two anyway.
Where camera analytics genuinely help
We are not going to pretend our software is a screening system. Here is the work it does do, all of it on cameras you have already paid for.
Everywhere there is no lane
You will put screening on one or two doors. Cameras already cover the garage, the loading dock, the perimeter fence, the side entrances, and the corridors. Visible-weapon detection runs on all of them at once.
The moment after screening fails
Screening is a filter, not a guarantee, and a weapon that gets past a lane or arrives through a staff door is then a visible weapon. That is exactly the condition camera AI is built for.
Outside the building
Most incidents that reach a doorway were visible in a parking lot first. Cameras watching approach routes buy the minutes that a portal at the threshold cannot.
Without adding headcount
Analytics do not need an attendant. They watch every camera continuously and only involve a human when something fires, which is the opposite of the staffing curve a screening lane puts you on.
Beyond weapons
The same deployment does intrusion, loitering, line crossing, vehicle and license plate detection, and plain-English search across recorded video, so the spend is not single purpose.
On your existing hardware
Any ONVIF or RTSP camera works. There is no appliance to buy, no camera replacement, and no rip and replace of a VMS you already run.
If you are working through a specific mandate, the requirements differ sharply by setting. We cover what California actually requires of hospitals on the hospital weapons detection systems page, including the compliance date most vendors get wrong, and the K-12 picture on weapons detection for schools. For the camera side of the question in isolation, can AI detect concealed weapons goes through the physics in more detail than we have room for here.
Questions buyers actually ask
What is a concealed weapons detection system?
It is a screening device at a controlled entrance that detects a weapon hidden on a person or in a bag, typically by sensing how metallic mass distorts a low-intensity electromagnetic field, then indicating where on the body the object sits. Unlike a camera, it senses through clothing and bags, which is what makes it able to find something concealed.
How does a weapons detector work?
A person walks between two sensor columns that generate a low-intensity field. Metallic and conductive objects disturb that field in patterns that vary by mass, shape, and orientation. Software compares the signature against profiles for threat items and everyday objects, then alarms and shows the operator a location on a body outline so secondary screening can be targeted rather than general.
Can AI cameras detect concealed weapons?
No. A firearm inside a bag, waistband, pocket, or under a coat gives a camera nothing to see, so no visible-light system can detect it. This is true of every camera vendor in the market, including us, ZeroEyes, Omnilert, Actuate, and Scylla. Camera AI detects a weapon once it is drawn and in view, which is a different job.
Can a concealed weapon be visible?
Sometimes, and that is the gap camera analytics cover. Printing through clothing, a grip showing above a waistband, or a weapon carried openly for the last stretch of an approach are all visible conditions. Camera AI is trained on weapons that are actually in view, so it acts on the moment a concealed weapon stops being concealed.
What is the difference between a weapons detection system and a metal detector?
A traditional metal detector alarms on metal, so visitors empty their pockets and pass single file. A modern screening portal tries to classify what the metal is, letting people walk through at pace carrying phones and keys. That is the promise. Whether a given product delivers it is exactly the claim the FTC required Evolv to stop asserting in November 2024.
How much does a concealed weapons detection system cost?
No vendor publishes a rate card. The most reliable public figures come from bid documents: the Buffalo Public Schools solicitation priced a single indoor lane at $78,810 over 48 months and a dual outdoor configuration at $153,290. Staffing costs more over time, since a 24/7 post takes roughly four to five full-time equivalents.
Is Motorola concealed weapons detection the same as Evolv?
It uses the same underlying technology. Motorola Solutions describes CWD as powered by Evolv Technology, and its 2021 announcement said the product came about through an agreement with Evolv Technologies. If you are comparing Motorola CWD against Evolv Express, you are largely comparing two packagings of one sensor platform, with different integrations and support around it.
Which concealed weapons detection vendor is most accurate?
Nobody can answer that honestly from public information, because no screening portal vendor publishes a detection accuracy rate you can verify or compare. Any ranking you find online is built on numbers the vendors did not publish. Judge these products on tested sensitivity settings at your own site, nuisance alarm rates in real traffic, and what the vendor will put in writing.
Do concealed weapons detection systems need staff?
Yes, and this is the cost that surprises people. Every lane needs an attendant whenever the entrance is open, plus staff to resolve alarms without creating a queue. Alarm resolution scales with visitor volume, not with the number of weapons found, so a busy entrance needs more people even in a year when it finds nothing.
Do I need entrance screening if I already have AI cameras?
If your requirement is to stop a hidden weapon at a door, yes, because cameras cannot do that. If your requirement is to know quickly when a weapon appears anywhere on your property, cameras cover far more ground for far less money. Most organizations with a real threat profile run screening at one or two doors and analytics everywhere else.
Related Reading
Motorola Concealed Weapons Detection Review
What CWD is, who actually builds it, and the claims to get in writing.
Can AI Detect Concealed Weapons?
The physical limit every camera vendor shares, explained properly.
Evolv Express Pricing
Real deployment costs from a signed public RFP rather than a rumored monthly rate.
Best Gun Detection Software
The camera AI vendors compared on what each publishes and what it omits.
AI Weapons Detection Software
Visible-weapon detection on existing cameras, and what it costs per camera.
Weapons Detection Systems for Hospitals
What California AB 2975 requires, and the deadline most vendors print wrong.
Cover Everywhere a Screening Lane Cannot Reach
Surveillant runs visible-weapon, person, vehicle, intrusion, and license plate detection on the ONVIF and RTSP cameras you already own, at a published $39 to $42 per camera per month, with a free plan for one camera. Start with your worst-lit exterior camera and see what it catches.
No credit card required. We do not sell screening portals and our software does not detect concealed weapons.