evidoria

← Back to browse

Good practice Imported

Evolv's AI Weapons Scanner Fails a Utica Stabbing Test, Then Draws an FTC Settlement

United States of America · Utica · See the United States of America profile

Evidence: Observational / pre–post Top 98% 13/100 · Ask Evidence Copilot about this practice

Utica, NY's $3.7M Evolv AI weapons scanner missed a knife used in an October 2022 stabbing; the FTC later ruled Evolv deceived schools about its AI's weapon-detection accuracy nationwide.

3.7 US$ million
Initial scanner contract cost
~250,000 US$
Cost to remove and replace after the incident
800+ schools, ~40 states
Schools deployed nationally
up to 60%
Independently cited false-alarm rate in some schools
~8%
Customer base eligible for FTC-mandated contract cancellation
Evolv's AI Weapons Scanner Fails a Utica Stabbing Test, Then Draws an FTC Settlement

Details

Maturity
Discontinued
Promoter
Evolv Technologies / Utica City School District / U.S. Federal Trade Commission
Period
2022-2024
Keywords
school safety, AI weapons detection, security screening, regulatory enforcement, K-12

Context

Thomas R. Proctor High School in Utica City School District, New York, installed Evolv Technologies' 'Evolv Express' AI-powered weapons-detection scanners for about $3.7 million, replacing traditional metal detectors on the promise that machine-learning models could reliably distinguish weapons from harmless items.

Activities

On 31 October 2022, a 17-year-old student carried a hunting-style knife through the scanner undetected and used it to stab another student. Utica administrators decommissioned the Evolv system and reverted to conventional metal detectors, spending an additional roughly $250,000 to remove and replace the equipment.

Results

Reporting documented similar detection failures and high false-alarm rates (independent testing cited rates up to 60% in some schools) at other U.S. school districts using the same technology, out of more than 800 schools across roughly 40 states where Evolv scanners had been deployed.

Conclusions

In November-December 2024 the FTC settled with Evolv, finding the company had deceived schools by claiming its AI could reliably detect 'all' weapons more accurately and cost-effectively than metal detectors, when internal and independent testing did not support those claims. The settlement barred unsupported AI-detection claims going forward and let a subset of K-12 customers (about 8% of Evolv's customer base) cancel remaining contracts, without a fine. The case is a rare instance of a national regulator formally sanctioning an AI vendor for exaggerating an education-safety product's AI capabilities, though the settlement's remedies were narrow and hundreds of schools reportedly continued using Evolv scanners afterward.

Implementation

Indicative cost
Medium (€50k–€500k) — $3.7 million initial deployment; ~$250,000 to remove and replace after the 2022 incident.
Time to results
Medium (1–3 years) — Deployed prior to the October 2022 incident; decommissioned in Utica shortly after; FTC settlement November-December 2024.
Staffing & skills
Utica City School District, Evolv Technologies (vendor), U.S. Federal Trade Commission (enforcement)

Common failure modes

  • AI scanner failed to detect a knife later used in a stabbing
  • vendor's marketed accuracy was not supported by internal or independent testing (FTC finding)
  • high false-alarm rates reported at other deploying schools
  • settlement remedies were narrow -- limited opt-out window, no fine -- and many schools continued using the product afterward

Where it fits

Governance type
school district with private AI-safety vendor; national regulator (FTC) intervention
Scale
single district incident with nationwide vendor deployment
Income level
high income (United States)

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful