The Evolv FTC Settlement and What It Means for Agencies Co-Deploying AI Weapons Detection

In late 2024, the Federal Trade Commission concluded that Evolv Technology, the largest vendor of AI-branded weapons-detection scanners in K-12 schools, had misled its customers about what its systems could actually do. A federal court approved the resulting settlement, and the case has kept surfacing since — most recently in local disputes over scanners that failed to catch weapons after purchase. For law enforcement agencies that advise school districts on security technology or co-deploy scanners alongside district staff, the settlement isn’t background news. It’s a live standard for what accuracy claims agencies should demand before putting their name behind a system.

What the FTC Actually Found

The FTC’s complaint, detailed in its own press release, alleged that Evolv marketed its Express scanners as more accurate than traditional metal detectors, capable of screening large volumes of people without slowing foot traffic and cutting security labor costs by 70 percent — claims the agency said weren’t supported by testing. The commission also found that Evolv’s own testing showed its systems detecting only metallic objects, despite branding built around “AI weapons detection,” and that the scanners generated high false-alarm rates on ordinary items like laptops, binders, and water bottles while missing actual weapons in documented incidents, including a case where an Express scanner failed to detect a seven-inch knife later used in a stabbing.

That incident is instructive beyond its own facts. After it happened, the affected school raised the scanner’s sensitivity to try to close the gap — and the FTC’s account of the episode notes the adjustment pushed the false-alarm rate to roughly 50 percent. That’s the tradeoff curve the marketing obscured: turning sensitivity up to catch more genuine threats also floods staff with false alarms at a rate that makes the system harder to operate meaningfully, and turning it back down reopens the original gap. A vendor claiming both high accuracy and low false-alarm rates simultaneously is describing a performance point the underlying sensor technology may not actually support.

What the Settlement Actually Requires

The approved settlement involved no monetary penalty and no admission of wrongdoing — a detail worth stating plainly, since it’s easy to read a settlement as a finding of liability when the practical outcome here was narrower. What it does require: Evolv is barred from claiming its systems detect weapons more accurately than metal detectors, or from making unsupported claims about AI-driven detection performance generally, and certain K-12 customers were given a 60-day window to cancel remaining contract terms. The order constrains future marketing. It does not retroactively validate or invalidate any specific deployment already in the field.

The Testing Question Agencies Actually Need Answered

For a police department or campus safety unit that recommends, co-deploys, or trains staff on a weapons-detection system, the settlement’s relevance isn’t “was Evolv specifically bad” — it’s that vendor-supplied accuracy figures for this entire product category have now been found, by a federal regulator, to sometimes reflect marketing rather than independent test results. That should change what agencies ask for before deployment, regardless of vendor:

• Independent testing data, not vendor-published detection rates, ideally from a lab or evaluation body with no commercial relationship to the manufacturer
• False-alarm rate under realistic conditions, not a controlled demo environment — the gap between demo performance and field performance is exactly where Evolv’s marketing diverged from its own internal data
• A clear description of what the sensor actually measures. A system marketed as “weapons detection” that is mechanically a metal detector with pattern-recognition software layered on top is a materially different product than true multi-sensor threat detection, and the settlement turns partly on that gap between branding and mechanism.

Vendor Neutrality in Practice

Agencies that advise on these procurements are not in a position to certify a vendor’s technology, and shouldn’t present themselves as doing so — but they are in a position to ask school boards and district security committees to require the kind of documentation the FTC found missing here before a contract is signed. That’s a meaningfully different role than endorsing a specific product, and it’s the one that holds up if a system underperforms after installation. Several districts using Evolv scanners have faced local scrutiny over exactly this gap between purchase and performance, which is the scenario a documentation requirement is meant to prevent before it becomes a public dispute.

Beyond Evolv: A Category-Wide Signal

Evolv held the largest share of the school weapons-detection market when the FTC acted, which is why this case reads as a company-specific story. The more useful reading treats it as a market signal: several competing vendors sell AI-branded screening systems using similar language about accuracy and reduced staffing needs, and none of that language has been independently validated at the standard the FTC applied here. An agency that only scrutinizes Evolv specifically, and takes a competitor’s marketing at face value because it isn’t the vendor named in a federal settlement, hasn’t actually closed the gap the case exposed — it’s applied the lesson to one company instead of one product category.

Agencies that sit on district security committees or safety task forces are also well positioned to push for a standard testing protocol before a purchase, rather than a vendor demo, which is inherently run under conditions the vendor controls. A demo that never floods the system with the ordinary volume of laptops, water bottles, and metal binder rings a real school day produces will not reveal the false-alarm behavior that mattered most in the Evolv case.

None of this means AI-assisted screening technology as a category is unreliable — it means the burden of proof for accuracy claims sits with independent testing, not vendor marketing, and the FTC’s finding here is now the reference point any agency can point to when asking a vendor to substantiate a number rather than accepting it. Agencies that build that requirement into procurement now are the ones least likely to be explaining a detection failure to a school board later.