AI Regulation Tracker / Enforcement and liability
The FTC Made a Company Unwind AI Weapons Claims and Let Customers Walk
A stipulated Section 5 order barred Evolv Technologies from unsupported claims that its AI detects weapons, and it forced a rare remedy: certain K-12 school customers got a 60-day window to cancel multi-year contracts. The complaint and stipulated order were filed November 26, 2024, and approved in early 2025. If you buy or sell "AI-powered" security or screening, this is the enforcement template to read.
Most AI enforcement so far has been about the claim: say your product is more accurate than it is, and the regulator makes you stop saying it. The FTC's action against Evolv Technologies went a step further. It made the seller stop making the claim, and it made the seller offer to unwind the sale. That second part is what makes this case worth your time even if you have never touched a weapons scanner in your life.
Evolv makes the Evolv Express, an AI-powered screening system that visitors walk through at entrances. According to the FTC's complaint, school systems make up half of Evolv's business, with Express scanners in over 800 schools across 40 states, plus stadiums and hospitals. The company marketed the product as a high-tech, AI-driven alternative to the ordinary metal detector. The FTC alleged that the marketing did not hold up.
What the FTC said Evolv claimed, and what it said was false
The FTC's complaint alleged that Evolv "deceptively advertised that its Evolv Express scanners would detect all weapons and made misleading claims that its use of artificial intelligence makes its screening systems more accurate, efficient, and cost-effective than traditional metal detectors." The agency listed the specific representations it said were unsupported: that the system will detect all weapons, ignore harmless personal items without making people empty their pockets or bags, detect weapons more accurately and faster than metal detectors, cut false alarms, and reduce labor costs by 70 percent.
The complaint backed the deception theory with operational failures, not just puffery. The FTC alleged that Express scanners "failed in several instances to detect weapons in schools while flagging harmless personal items typically brought to schools, like laptops, binders, and water bottles." Its most pointed example: a seven-inch knife brought into a school in October 2022 that the system did not catch and that was used to stab a student. After that, the FTC said, the school raised the sensitivity and hit a 50 percent false alarm rate. In other words, the "AI" edge the marketing promised collapsed into the same tradeoff any metal detector forces, more sensitivity for more false alarms.
Samuel Levine, then Director of the FTC's Bureau of Consumer Protection, framed the standard the agency was applying: claims about technology "including artificial intelligence" need to be backed up, "and that is especially important when these claims involve the safety of children." That is the whole doctrine in one line. AI is not a magic word that lowers your burden of proof. If anything, invoking it raises the stakes.
The claims ban, and why the wording matters
The order prohibits Evolv from misrepresenting the ability of its products to detect weapons or ignore harmless items, their accuracy and false alarm rates compared with metal detectors, screening speed, labor costs, and testing results. The catch-all is the part to underline. Evolv is barred from misrepresenting "any material aspect of its performance, including the use of algorithms, artificial intelligence, or other automated systems or tools."
Read that as a drafting instruction for your own marketing. The FTC did not carve out AI as some special, softer category of claim. It folded "artificial intelligence" into the ordinary law of material performance representations. A claim about what your model does is a performance claim, and it needs the same substantiation you would need for any other objective product claim. If you cannot support it with competent and reliable evidence, do not make it, and do not launder it through the word "AI."
The unusual part: a contract-unwind remedy
What sets this order apart from a routine advertising settlement is the cancellation remedy. In the FTC's own description, Evolv "would also have to give certain K-12 school customers the option to cancel their contracts, which generally lock customers into multi-year deals." The order required Evolv to notify eligible schools, defined as K-12 customers that purchased or contracted for Evolv Express between April 1, 2022 and June 30, 2023, that they could cancel. Eligible customers had 60 days from receiving the notice to opt out, and they could choose whether cancellation took effect immediately or at a natural break like the end of a term.
This is rare. Advertising cases usually end in a claims ban and sometimes money. Forcing a seller to release customers from signed, multi-year contracts is a different order of remedy, one that reaches the deal itself rather than just the pitch. Evolv later reported that a large majority of eligible schools chose to stay, which is its own useful data point, but the remedy is the precedent. The FTC showed it will pursue contract unwinding where it believes the sale was built on deceptive AI performance claims and customers are locked in.
What this means for US buyers and sellers of "AI-powered" anything
If you sell AI-enabled products in the United States, treat this order as the FTC's published playbook. Every objective claim about detection, accuracy, false positives, speed, or cost savings is a performance claim that needs substantiation before it goes in a deck, a data sheet, or a sales call. "Our AI is more accurate" is not marketing color, it is a factual representation the FTC can and will test against how the product performs in the field. Keep the testing that supports each claim, and make sure the claim matches the tested configuration, not the ideal one.
If you buy these systems, the lesson is on the other side of the table. The FTC alleged that Evolv's own guidance to schools quietly walked back the marketing, advising more staffing, conveyor belts, and hand-checks that made the system behave more like the cheaper metal detector it was sold to replace. Before you sign a multi-year deal on an AI performance promise, get the vendor's real-world detection and false alarm figures in writing, tied to your deployment, and put the claims that sold you into the contract. Procurement and counsel should also note that the deception here reached the enforceability of the deal itself. A sales process built on unsupported AI claims is not just a marketing risk, it is a contract risk.
Questions professionals are asking
Is this a binding order or just an FTC statement?
It is binding on Evolv. The resolution is a stipulated order for a permanent injunction, meaning the company agreed to it, entered by a federal court in the District of Massachusetts under Section 5 of the FTC Act. It is company-specific enforcement rather than a rule that binds other firms, but it states the substantiation standard the FTC applies to AI product claims.
What exactly is Evolv barred from claiming?
The order prohibits misrepresentations about the products' ability to detect weapons and ignore harmless items, their accuracy and false alarm rates versus metal detectors, screening speed, labor costs, testing results, and "any material aspect of its performance, including the use of algorithms, artificial intelligence, or other automated systems or tools." AI claims are treated as ordinary material performance claims that require evidence.
What was the contract cancellation remedy?
Evolv had to notify certain K-12 school customers that they could cancel their contracts. Eligible schools were those that bought or contracted for Evolv Express between April 1, 2022 and June 30, 2023, and they had 60 days from receiving the notice to opt out of otherwise multi-year deals, effective immediately or at a convenient point such as the end of a term.
Why does this matter if I do not sell weapons scanners?
Because the doctrine is general. The FTC treated "AI" as an ordinary performance claim requiring substantiation, and it reached the contract, not just the advertising. Any US seller or buyer of products marketed as "AI-powered," in security or anywhere else, should read it as the enforcement template for AI accuracy claims.
What should buyers of AI security systems do now?
Get the vendor's real-world detection and false alarm figures in writing, tied to your specific deployment, and put the AI performance claims that sold you into the contract itself. Have counsel review multi-year commitments, since this case shows deceptive AI claims can affect the enforceability of the deal, not only the marketing.
RELATED BRIEFINGS
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal advice. Confirm how any order or requirement applies to your situation with qualified counsel in the relevant jurisdiction.