“Creepy? Sure. Great for marketing? Definitely.”
That line, according to the US Federal Trade Commission, appeared on a Cox Media Group website promoting a service called Active Listening. The pitch to small businesses was that artificial intelligence could pick out people talking about a need, say a leaking roof or a new car, from conversations overheard by their phones, smart TVs and smart speakers, and then serve those people ads at exactly the right moment (FTC, 2026c, paras. 4–6).
It was the kind of claim that confirms a fear many of us already have. It was also, the FTC says, untrue. The service “did not collect or use voice data in any manner” (FTC, 2026c, para. 9). In August, the FTC finalised orders requiring Cox Media Group and two partner firms to pay a combined $930,000 and to stop misrepresenting what their services do (FTC, 2026b).
This is a story about a lie, but not the one you might expect. Nobody’s phone was secretly recording them for this service. The people misled were the small businesses that paid for it, and the thing that sold the lie was the word “AI.”
What the businesses were told
Cox Media Group, a large US media company operating as CMG Media Corporation, sells advertising and marketing services to small businesses (FTC, 2026c, para. 2). Beginning in 2023, it offered Active Listening to those customers through a deal with a firm called MindSift, which sold the service on a white-label basis (para. 4). A third company, 1010 Digital Works, worked with MindSift to offer it directly to small businesses and to CMG (FTC, 2026d, paras. 2, 4).
The FTC’s complaint quotes the marketing at length. The website told business owners that “your smartphone is technically always listening” and promised that “every casual conversation between two consumers becomes a tool for you to target, retarget, and retain customers.” It went on: “It may seem like black magic, but it’s not—it’s AI” (FTC, 2026c, para. 5). Businesses could, it said, claim exclusive territories in 10- or 20-mile radiuses.
In sales presentations from 2023 to mid-2024, the complaint says, CMG described AI analysing conversations “via smartphones, smart tvs, smart speakers and other devices” (para. 6). Sales staff asked prospects, “Where do you want us to listen?” When some customers were sceptical, the FTC alleges, CMG “doubled down,” naming well-known companies as among “570 different data sources” that supplied voice data (para. 7). 1010 Digital Works’ own pitch, the FTC says, was blunter still: “Yes, your devices are listening to you” (FTC, 2026d, para. 5).
When customers asked whether this was even legal, the answer, according to the complaint, was that consumers had “opted in” through the terms of service they accept when they download apps or set up new devices (FTC, 2026c, para. 8).
What the FTC says was really happening
According to the FTC, none of it was true. Smart devices did not send voice data to the service, and there was no AI detecting conversations. Instead, the complaint says, the service was “nothing more than consumer email list buying”: lists of email addresses bought from data brokers, of people presumed to have certain interests, which CMG resold “at a significant markup” (FTC, 2026c, para. 9).
The local targeting did not hold up either. A business that paid for, say, a ten-mile radius around Orlando, Florida, got lists drawn “from across the country, with only a fraction” of the people anywhere nearby (para. 11). And the FTC says no consumer had consented to having their voice data collected for marketing, because there was no such collection to consent to (para. 10).
These are the FTC’s allegations. The companies settled without admitting them: CMG’s order records that it “neither admits nor denies” the allegations in the complaint (FTC, 2026a, p. 1).
Key idea: AI-washing
“AI-washing” means overstating, or inventing, what artificial intelligence does in a product, because the label sells. It matters in two directions. Buyers pay for capabilities that do not exist and make decisions based on them. And the public is left with a distorted picture of what technology can do, which makes real risks harder to see. The fix is simple to say and harder to do: ask what data the system actually uses, and ask for evidence, not adjectives.
What the orders require
The FTC announced the cases in May 2026, when the Commission voted 2–0 to issue the complaints and accept proposed settlements. “It is a basic rule of business that you need to be honest with your customers, and these companies failed to do that,” said Christopher Mufarrige, director of the FTC’s Bureau of Consumer Protection (FTC, 2026e). After a public comment period that drew two comments, the Commission voted 2–0 to make the orders final (FTC, 2026b).
Under the final orders, CMG pays $880,000, and MindSift and 1010 Digital Works pay $25,000 each. The FTC says the money is for redress to CMG’s affected customers (FTC, 2026b). Each company is barred from misrepresenting the qualities or features of its marketing services, its collection and use of voice data, whether consumers consented, and its ability to target ads geographically (FTC, 2026a, Provision I). CMG’s order runs for 20 years (Provision IX).
The FTC also made a point worth underlining. Even if the service had worked as advertised, collecting and using people’s voice data without adequate consent “would itself violate the FTC Act” (FTC, 2026b). In other words, a line buried in an app’s terms of service is not the kind of consent the FTC had in mind. The “they opted in” defence failed twice: the data did not exist, and the consent would not have counted.
Why a fake AI claim is a real problem
It would be easy to file this under marketing hype and move on. We think that would miss the point.
First, the people harmed were small businesses: roofers, dentists, car dealers, whoever bought a promise of knowing “exactly when someone is in the market” (FTC, 2026c, para. 5). Small firms rarely have the technical staff to test a vendor’s claims. They rely on the vendor telling the truth.
Second, the word “AI” did real work in the sale. It made an implausible claim sound plausible, and it made scepticism sound out of date. That is why regulators have been treating inflated AI claims as ordinary deception. In 2024, under a different chair, the FTC launched Operation AI Comply against deceptive AI schemes, with then-Chair Lina Khan stating that “using AI tools to trick, mislead, or defraud people is illegal” (FTC, 2024). The Cox Media case, brought under the current leadership, shows the principle has outlasted the change of administration.
Third, the story feeds a public confusion that cuts both ways. If you believed your phone was listening for advertisers, this pitch seemed to confirm it. In fact the record shows the opposite for this service: no voice data was involved. But that does not prove that no company anywhere collects voice data, and the FTC’s point about consent shows why it would matter if one did. The honest takeaway is narrower and more useful: claims about what AI does need evidence, whether they come from a vendor or from a viral post.
What the record does not show
The record does not tell us how many businesses bought Active Listening, how much they paid in total, or how much each will receive in redress. The FTC’s findings are allegations that the companies neither admitted nor denied; no court tested them. We did not open MindSift’s separate complaint, so its role here is described from the other two complaints and the FTC’s announcements. And the case says nothing, either way, about whether other apps or devices listen to conversations for advertising. It is about one service whose sellers said it did, when, according to the FTC, it did not.
Questions you can ask
- If you are worried about your devices: check which apps have microphone access in your phone’s privacy settings, and switch it off for any that do not need it. Agreeing to terms of service is not, in the FTC’s view, the same as meaningful consent to voice data collection.
- If you run a small business buying marketing or “AI” services: ask the vendor exactly what data the service uses, where it comes from, and how you could check the results, for example whether the leads are really in your area.
- If you buy AI for an organisation: ask for evidence that the AI feature exists and does what is claimed, such as test results, documentation or a trial on your own data, and write the vendor’s claims into the contract.
- If you sell AI products: make sure every claim in your marketing and sales scripts can be backed by a record you could show a regulator today.
Go further. Operational AI Readiness is FairByDesign’s field manual for AI that is already live, and the complete Standard edition is free to read and download. Chapter 11, What Auditors, Regulators, and Internal Reviewers Will Actually Look For, covers why a claim without a record behind it does not survive scrutiny, and Appendix D, the vendor due-diligence question bank, gives you the questions to put to anyone selling you “AI.” Read it free, or test one workflow with the free readiness assessment, paying close attention to domain 1, inventory and scope, and domain 4, the evidence spine.
Cases like this are easy to misread from a headline. Reading the complaints and orders, and separating what was alleged, what was settled and what is still unknown, is what FairByDesign does. We do it without advertising and without money from the companies we write about.
Help keep independent scrutiny going. Hype about AI is cheap; checking it against the record takes time. FairByDesign is reader-funded and takes no money from the companies we examine, so your support is what keeps that line clear. Give once ($5, $10, $25, $50 or any amount), give monthly ($10 or $25), or become a patron: fairbydesign.org/patrons-2. Support never buys editorial control.
Not in a position to give? The book is free for everyone: read Operational AI Readiness.
References
Bakos, V. (2026). Operational AI readiness: A practitioner guide to governing AI that is already live. FairByDesign. https://fairbydesign.org/operational-ai-readiness/
Federal Trade Commission. (2024, September 25). FTC announces crackdown on deceptive AI claims and schemes [Press release]. https://www.ftc.gov/news-events/news/press-releases/2024/09/ftc-announces-crackdown-deceptive-ai-claims-schemes
Federal Trade Commission. (2026a, August 26). In the matter of CMG Media Corporation: Decision and order (FTC Matter No. 242-3029). https://www.ftc.gov/system/files/ftc_gov/pdf/2423029c4838cmgfinalorder.pdf
Federal Trade Commission. (2026b, August 27). FTC finalizes orders with Cox Media Group, two other firms settling charges they deceived customers about “Active Listening” AI-powered marketing service [Press release]. https://www.ftc.gov/news-events/news/press-releases/2026/08/ftc-finalizes-orders-cox-media-group-two-other-firms-settling-charges-they-deceived-customers-about
Federal Trade Commission. (2026c). In the matter of CMG Media Corporation: Complaint (FTC Matter No. 242-3029). https://www.ftc.gov/system/files/ftc_gov/pdf/CMGComplaintwithoutsignatures.pdf
Federal Trade Commission. (2026d). In the matter of 1010 Digital Works LLC: Complaint (FTC Matter No. 242-3033). https://www.ftc.gov/system/files/ftc_gov/pdf/1010digitalworksllccomplaintwithoutsignatures.pdf
Federal Trade Commission. (2026e, May 21). FTC to require Cox Media Group, two other firms to pay nearly $1 million to settle charges they deceived customers about “Active Listening” AI-powered marketing service [Press release]. https://www.ftc.gov/news-events/news/press-releases/2026/05/ftc-require-cox-media-group-two-other-firms-pay-nearly-1-million-settle-charges-they-deceived
Leave a Reply
You must be logged in to post a comment.