In October 2022, a student walked into a New York high school carrying a seven-inch knife. The school had paid for a scanner sold as AI weapons detection technology. The scanner did not catch the knife, and another student was stabbed. Two years later, that incident appeared in a federal case brought by the Federal Trade Commission.
The scanner at the center of that case is used in thousands of schools, stadiums, and other venues where screening large crowds quickly is a priority. According to the FTC, schools make up about half of the vendor’s business, with its scanners in more than 800 schools across 40 states.
At Garrett, we don’t see this case as a verdict against AI. We see it as a lesson in where AI belongs in detection. In our view, AI is a powerful tool for engineers to design, test, and improve a detector before it ships. It should not be a promise that changes behavior at the school entrance. The findings described here are what the FTC alleged, and the company neither admitted nor denied the allegations when it settled.
Key Takeaways
- In November 2024, the FTC sued a leading AI weapons screening company, alleging it made unsupported claims about its walk-through scanners, and the company settled without admitting or denying the allegations.
- The FTC said the “weapons detection” label was only a marketing distinction, because the scanners detect metal much like traditional metal detectors.
- A scanner at a Utica, New York school reportedly missed a seven-inch knife that was later used to stab a student.
- Raising the sensitivity afterward pushed false alarms to about 50%, which erased the frictionless entry that was the product’s main selling point.
- Before signing any contract, schools should ask vendors for independent test results, expected false alarm rates, and clear proof of the ability to detect the school’s smallest threat objects.
- Garrett uses AI only during design and testing, then locks the software so its detectors perform the same way every day.
The Case at a Glance
Key Dates and Filings
On November 26, 2024, the FTC filed suit against the company in federal court in Massachusetts. The Commission voted 5-0 to authorize the complaint and proposed a stipulated order. The case was part of Operation AI Comply, an FTC effort to make sure companies can back up the AI claims in their marketing.
On December 5, 2024, the court entered the stipulated order, which made the settlement terms binding. The table below sums up the main facts from the public record.
| Item | Detail |
| Date filed | November 26, 2024 |
| Court | U.S. District Court, District of Massachusetts |
| Case number | 1:24-cv-12940 |
| Commission vote | 5-0 to authorize complaint and order |
| Order entered | December 5, 2024 |
| Broader effort | Operation AI Comply |
| Admission of wrongdoing | None. The company neither admitted nor denied the allegations |
| Main remedy | Contract cancellation offer for certain K-12 customers |
The Contract Cancellation Remedy
Under the order, the company had to tell certain K-12 customers they could cancel contracts signed between April 1, 2022, and June 30, 2023. Schools with larger contracts of 15 or more units were not eligible. As IPVM noted in its report on the complaint, that left out many of the biggest districts using the product.
In its January 2025 business update, the company said no customers had used the cancellation right as of January 23, 2025. Most of those rights expired by March 14, 2025. Some readers see that as a sign of customer loyalty, while others see it as a sign that contract terms made cancellation hard in practice. Either way, contract length deserves a place in your long-term security planning.
What the FTC Alleged
The Five Challenged Claims
The FTC alleged that the company made claims about its scanner that it could not support. According to the complaint, the company misrepresented that its scanner would:
- Detect all weapons
- Ignore harmless personal items without people emptying pockets or bags
- Outperform metal detectors on accuracy and speed
- Lower false alarm rates
- Cut labor costs by 70% compared with metal detectors
Each of these claims speaks to what school leaders care about most. Safety, speed, fewer hassles, and lower staffing costs are exactly what a school security budget is meant to buy. That is why the gap between the pitch and the alleged performance matters so much.
The “Weapons Detection” Label
The complaint said that calling the scanner “weapons detection” instead of “metal detection” was solely a marketing distinction. The FTC alleged the scanners detect only metallic objects and can alarm on metal items that are not weapons at all.
This point matters for buyers. If the basic physics is the same as magnetometer security screening, then the label alone tells you nothing. The real comparison comes down to three things: performance, transparency, and cost.
Testing Claims Under Scrutiny
The complaint also questioned how the company described third-party testing of its product. According to then-Commissioner Andrew Ferguson’s concurring statement, the FTC alleged that the company asked to have negative efficacy information removed from a testing report.
Ferguson’s summary also noted that the company sponsored luncheons at the testing organization’s conferences. For schools, the takeaway is simple. A test summary in a sales deck is not the same as a full, independent report.
The Failure That Made Headlines
A Missed Seven-Inch Knife
The AI scanner reportedly failed to detect a seven-inch knife brought into a school in October 2022. The knife was then used to stab a student. The complaint references this stabbing in Utica, New York.
This was not a lab result or a test score. It was a real failure at a real school entrance. Incidents like this show why claims about stopping weapons before violence need proof behind them.
The Fix That Undercut the Pitch
After the stabbing, officials raised the scanner’s sensitivity. This produced a false alarm rate of about 50%. Half of the people walking through were being flagged. Here is the key point. Frictionless entry was the main selling point. That advantage disappeared the moment safety became the top priority.
The Core Problem: “AI” as a Marketing Promise
The Settings Still Decided the Outcome
The issue wasn’t that AI let weapons through. It was that “AI” was used as marketing spin, presenting the scanner as a black-box wonder tool that could handle detection on its own. Schools came away believing the technology would sort out the hard tradeoffs for them. It didn’t. Like any screening system, it still had to be configured for what a school was trying to catch, and that responsibility stayed with the school.
The complaint alleged that the scanner did not reliably detect guns and knives at average sensitivity. Schools could raise the setting to catch more weapons, but that brought more false alarms, and every alarm meant stopping a student for a secondary check.
Better security came at the direct expense of the vendor’s headline claim: fast, high-volume throughput. The FTC cited a higher false alarm rate at sensitivity setting “G,” and IPVM’s reporting on the scanner’s sensitivity settings points to the same tradeoff. AI couldn’t deliver both strong detection and speed at once, but the marketing talked as if it could.
The Vendor’s Own Guidance Confirmed It
In 2023, the company released a higher-sensitivity setting aimed at knives. According to the complaint, the company admitted some knives would still be missed, false alarms would rise, and more staff might be needed.
The company also recommended conveyor belts and having people hand items around the scanner. The complaint said this made the scanner work more like lower-cost traditional metal detectors. At that point, the premium price was hard to justify.
Sensitivity Settings Are Not Unique, but Transparency Is
In a statement distributed on Business Wire, the company argued that sensitivity settings are not unique to its product. Traditional walk-through metal detectors have them too. That is a fair point.
But the scanner was marketed as an AI weapons detector that had solved the speed-versus-detection problem. Once you factor in sensitivity and false alarms, it’s back with the pack: its faster throughput comes partly from lower sensitivity and weaker response to non-ferrous threats. That’s why settings matter. A school should know what its scanner ignores and how target filtering works, with those settings documented and protected from casual changes.
At Garrett, We Have a Different Philosophy: AI in the Lab, Not on the Front Line
Where We Use AI
At Garrett, our policy is to use AI during the design phase. Once the work is done, we lock down the software so it performs the same way every time. Our Clarity platform, announced in January 2026, is a proprietary deep learning development platform that has already helped us improve the Garrett Paragon’s detection and indication.
This approach has a track record. Our September 2025 machine learning update improved the Paragon’s multi-target alarm indication and pinpointing. It did so without ghosting, zone drift, or over-alarming.
The Safeguards Behind Our AI Work
Using AI responsibly takes rules, not just good intentions. Our AI Policy lays out the safeguards we follow:
- Extensive testing of AI-generated IP and techniques
- Adequate testing of all AI work products before release
- No customer data or sensitive security information sent to off-site AI platforms
- NDAA-compliant cameras in our products to protect data interpreted by AI-built algorithms
These steps keep AI where we believe it belongs. It helps our engineers build better detectors. It does not make live, unchecked decisions at a school door.
Why “Locked Down” Matters for Schools
A locked system gives predictable performance from day one through year five. Your staff do not have to wonder if an algorithm in the detector could cause performance drift. Training and procedures stay valid over time.
Just as important, we prove AI gains before a product ships. They are not discovered by students walking through the door. Readers who want our full framework can review our AI Policy.
Transparency You Can Verify: What a Properly Programmed Garrett Paragon Offers
Control Without Guesswork
We built the Garrett Paragon to give schools detailed control over how screening works. The table below shows the main features and why each one matters.
| Feature | What the Garrett Paragon Offers | Why It Matters for Schools |
| Program & sensitivity | 200 levels, 23+ programs, 4+ school-specific programs | Fine-tune for your threat profile |
| Detection zones | 66 zones across 22 independently adjustable horizontal bands | Pinpoints where metal is on the body |
| Quick-Q settings | Filters cell phones while detecting mass casualty weapons | Fewer nuisance alarms with a visible threshold |
| Access control | Three levels of protected settings | Stops unauthorized changes |
| K-12 kit | NFC card set for touchless, consistent setup and management | Simple configuration for school staff |
Our Quick-Q settings take on the same nuisance alarm problem that “AI screening” promises to solve, including how phones affect screening. The difference is that we do not hide the threshold. Your staff know what the unit is looking for and what it is set to ignore.
Settings That Stay Where You Put Them
The Garrett Paragon’s settings are tamperproof, with three levels of access control. Only authorized people can make changes, so no one can adjust sensitivity or detection settings without the right permissions. Our Garrett Paragon kit includes an NFC card set. It makes setup and management easier for school teams who are not full-time security technicians.
Tamperproof settings stop unauthorized changes, but they don’t address AI drift: the gradual shift in how an AI model performs over time, even when no one touches it. That’s why our AI policy works alongside Paragon’s access controls. Access controls keep the configuration locked, and our AI policy keeps detection accurate, so your screening stays consistent from the first day of school to the last.
Five Questions to Ask Any Screening Vendor
The FTC case offers a clear checklist for buyers. Before signing any contract, school leaders should ask every vendor these questions:
- What is detection performance at the exact setting we will run, against the items we care about?
- What false alarm rate should we expect with typical student belongings?
- Who can change the settings, and how is that access controlled?
- If you use AI, is it validated and locked before deployment, or does behavior change in the field?
- Can we see full, independent test results, not marketing summaries?
A good vendor will answer all five in writing and explain what weapon detectors catch at each setting. If a vendor dodges any of them, treat that as an answer too.
Conclusion: Is Faster Throughput Worth Regulator-Documented Accuracy Issues?
The choice in front of school leaders weighing metal detectors in schools comes down to proof. On one side is a promise of faster entry. On the other is the fact that the AI scanner’s accuracy problems, as the FTC alleged them, are now part of a federal record. For students and staff, the cost of a missed weapon is far higher than a few extra seconds at the door.
At Garrett, we believe AI belongs in rigorous engineering, not as an unproven promise at the school entrance. We use it to design and test better detectors, then lock the software so our security detectors perform the same way on day one and year five. Our settings are clear, protected, and backed by testing you can review.
So, which matters more to you: a bold claim or a result you can verify? Contact our team at Garrett to learn more about our security detectors and find the right screening solution for your school.
