Evolv Weapons Detection vs AI Video Weapon Detection: How to Choose a Weapons Screening System
Two products carry the phrase "AI weapons detection" and they do not do the same job. One screens bodies at a doorway for concealed weapons. The other watches every camera you own for a weapon in someone's hand. Districts and facilities that confuse them buy the wrong system, and the federal record now contains an expensive example of what happens next.
Surveillant is an AI video analytics platform. It is not a screening portal, and this guide says plainly where a portal wins.
What Is the Difference Between Evolv and AI Video Weapon Detection?
Evolv Express is a weapons screening portal. People walk through it at a controlled entrance and sensors look for the signature of a concealed weapon on the body or in a bag. AI video weapon detection, the category Surveillant belongs to, runs computer vision on your existing camera feeds and raises an alert when a weapon becomes visible, anywhere a camera is pointed.
The practical difference is where the threat is caught. A portal catches a gun in a backpack before it enters the building, but only at the doorway that has a portal. Video detection catches a gun in a hand, in a parking lot, on a loading dock, or in a hallway with no portal in sight, but it cannot see through fabric.
So the question is not which product is better. It is which failure you cannot tolerate. A stadium with one controlled entry and a bag policy should screen. An open school campus with twelve doors, a bus loop, and a parking lot cannot screen all of it, and the weapon that matters is usually already drawn by the time it reaches a camera. Most large sites eventually run both.
Pick by threat model:
- Concealed weapon entering a building. You need screening.
- Weapon drawn anywhere on the property. You need video detection.
- Both, on a real campus. You need both, and you should budget for both.
Weapons Screening vs AI Video Weapon Detection
Evolv is the best-known name in screening, so it stands in for the modality here. The comparison holds for any walk-through screening vendor. Where screening wins, the table says so.
| Factor | Weapons screening (Evolv Express and similar) | AI video weapon detection (Surveillant and similar) |
|---|---|---|
| Concealed weapon in a bag or waistband | Detected. This is the whole point. | Not detected, ever |
| Weapon drawn in a parking lot | Not covered | Detected if a camera sees it |
| Coverage footprint | One entrance lane per unit | Every camera on the property |
| Hardware to buy | Portal units, plus queue space | None, uses existing IP cameras |
| Staffing to operate | Officers at each lane during entry | One reviewer to verify alerts |
| Effect on entry flow | Creates a queue at every screened door | None |
| Typical false positives | Laptops, Chromebooks, binders, umbrellas, water bottles | Phones, tools, umbrellas, toy guns |
| Cost shape | Capital purchase or lease per lane, plus maintenance | Software subscription per camera per month |
| Works after hours on an empty site | No, nobody is walking through | Yes, cameras keep watching |
| Best fit | Stadiums, arenas, courthouses, single controlled entry | Open campuses, retail, lots, multi-entrance sites |
The FTC Case, and Why It Should Change How You Buy
In November 2024 the Federal Trade Commission announced an action against Evolv Technologies, alleging the company made unsupported claims about its AI-powered screening system. According to the FTC's complaint, Evolv represented that Evolv Express would detect all weapons, ignore harmless personal items without requiring people to empty their pockets or bags, outperform metal detectors on speed and accuracy, reduce false alarms, and cut labor costs by 70 percent.
The proposed settlement bars Evolv from making unsupported AI weapons-detection claims and requires the company to let certain K-12 school customers that signed contracts between April 1, 2022 and June 30, 2023 cancel them. At the time of the complaint, the FTC said Express scanners were in place at more than 800 schools across 40 states. The scrutiny followed two school stabbings, in Ohio and New York, at buildings where Evolv scanners were installed and the knives were not detected.
None of that means screening is a bad technology. It means the accuracy number in a sales deck is not evidence. The useful lesson generalizes to every vendor in this category, including us: a detection claim that is not tied to a specific weapon type, at a specific distance, under specific lighting, on your cameras, is not a claim you can rely on.
So do this instead. Ask for a pilot on your own site, define in writing which weapons you expect to be caught and where, count the false positives per day for two weeks, and make the contract contingent on those numbers. Any vendor unwilling to be measured that way is telling you something. Our own guide to how accurate AI video analytics really is explains why a single percentage figure on surveillance video is close to meaningless.
A four-step pilot that produces real numbers
- Write down the threat you are buying against, in one sentence, before you see a demo.
- Run the system on your cameras or your doorway for two full weeks, in the seasons and lighting you actually have.
- Count every alert. Separate true detections from false ones, and log what caused each false one.
- Stage a supervised test with an inert training weapon at the distances that matter. Record how often it is caught, and where the misses cluster.
If the vendor will not allow step 4, the pilot cannot tell you what you need to know.
What Does Each One Cost?
Screening and video detection have different cost shapes, which is why they are hard to compare on a single line. Screening cost scales with the number of entrances you cover and the officers who staff them. Video detection cost scales with the number of cameras.
Reported figures for school weapons-detection installations put a basic system somewhere between roughly $10,000 and $50,000, with large facilities running past $200,000, and annual maintenance and software adding on the order of 10 to 20 percent of the initial spend. Baltimore City Schools approved more than $5 million to deploy across their buildings. These are third-party reported ranges, not quotes, and they vary enormously with the number of lanes.
AI video weapon detection has no portal to buy. It runs as an analytics subscription against cameras you already own. Analytics add-ons layered onto an existing system typically run $3 to $15 per camera per month; a full cloud platform including detection, alerting, and search runs $39 to $42 per camera per month with Surveillant. A forty-camera campus therefore lands in a very different budget line than a four-lane screening deployment, and it covers ground no portal reaches.
The trap in both cases is the second year. Screening carries maintenance, consumables, and the standing labor of officers at each lane. Video carries the subscription. Build a three-year total, not a purchase price, the way you would for any AI video analytics ROI case.
Weapons Detection Questions
What is Evolv weapons detection?
Evolv Express is a walk-through weapons screening system made by Evolv Technologies. People pass through a portal at a controlled entrance and sensors look for the signature of a concealed weapon on the body or in a bag, without requiring pockets to be emptied. It is widely deployed in US stadiums, hospitals, and schools, and it is a screening product rather than a camera analytics product.
Is there an alternative to Evolv weapons detection?
For concealed-weapon screening at a doorway, the alternatives are other screening vendors and conventional metal detectors, because the physics is the same. For weapon detection everywhere else on a property, the alternative is AI video weapon detection running on existing security cameras. It is not a substitute for screening; it covers the areas a portal cannot and catches a weapon once it is visible.
Did the FTC take action against Evolv?
Yes. In November 2024 the FTC announced an action alleging Evolv made false or unsupported claims that its AI screening would detect all weapons, ignore harmless items, outperform metal detectors, cut false alarms, and reduce labor costs by 70 percent. The proposed settlement bars unsupported AI detection claims and lets certain K-12 customers who contracted between April 2022 and June 2023 cancel.
Can AI cameras detect concealed weapons?
No. A camera records visible light, so a weapon inside a bag, coat, or waistband produces no image to detect. Any vendor selling camera software as concealed-weapon detection is claiming something the sensor cannot physically do. Concealed detection requires a screening modality that senses metal or composite material through fabric, which is what a walk-through portal is built to do.
Which is better for schools, weapons screening or AI cameras?
They answer different threats, so the honest answer is that it depends on the campus. A single-entry building with staff to run a lane benefits from screening at the door. An open campus with many entrances, a bus loop, and parking lots cannot screen its perimeter, and video detection covers that ground. Districts with budget for one should map their entrances first, then decide.
How many false alarms do weapons detection systems produce?
Enough that human verification is standard practice in both modalities. Reporting on school screening deployments has documented ordinary objects such as laptops, Chromebooks, binders, and water bottles triggering alerts at high rates, with one account describing roughly 30 percent of entries generating an alert. AI video detection produces its own false positives on phones, tools, and umbrellas.
Has AI weapon detection ever prevented a school shooting?
There is no published empirical evidence that any AI weapons detection system, screening or video, has prevented a school shooting. That absence is not proof the technology is useless, but it does mean the purchase should be justified by measurable detection performance on your site and by faster response, not by a prevention claim no vendor can support.
Does AI weapon detection work with existing security cameras?
Yes, if the cameras stream over RTSP or ONVIF and the image resolves a weapon at the distance you care about. Surveillant connects to those feeds directly, with no new hardware. The common constraint is placement rather than compatibility: cameras aimed at property lines for general security often frame people too small for a reliable weapon classification.
Related Reading
Test weapon detection on your own cameras, with your own numbers
Connect a few streams and measure detections and false positives for two weeks before you commit to anything. No credit card required.