FTC locks Cox Media Group's fake Active Listening ad product under 20 years of binding orders

The Federal Trade Commission issued final decision and orders on 26 August 2026 against CMG Media Corporation, MindSift LLC and 1010 Digital Works LLC, turning a $930,000 settlement into obligations that run for two decades. The case concerns Active Listening, an advertising product sold to small businesses as an AI system that detected purchase intent from conversations picked up by smartphones, smart TVs and smart speakers. According to the orders, no such capability existed.

The Commission announced the finalisation on 27 August 2026, three months after publishing the proposed consent agreements. PPC Land reported the details, including what buyers of Active Listening actually received in place of the conversation-based audio targeting they were sold.

What Active Listening was sold as

The product, marketed under the Cox Media Group banner, was pitched to small businesses as artificial intelligence that listened. The claim was that conversations captured by consumer devices, specifically smartphones, smart TVs and smart speakers, could be mined for purchase intent, and that ads could then be targeted at those people within a local radius.

That pitch works because it lands on a belief a very large number of people already hold. Almost everyone has had the experience of discussing a product out loud and then seeing an advert for it, and has concluded that the phone was listening. A sales deck that offers to sell you the thing you already suspect exists does not have to overcome much scepticism.

What buyers actually received

According to the FTC, the capability did not exist. What was delivered instead were consumer email lists purchased from third-party data brokers and resold at a markup. The geographic coverage of those lists bore little relation to the territories the businesses had paid for.

That mechanism is worth stating plainly, because it is the ordinary version of the fraud rather than the exotic one. Nobody built a surveillance system that failed. A broker list was bought, marked up, and described as something else. The gap between the pitch and the deliverable is not a matter of degree, it is a different product entirely, and the geographic mismatch means even the list itself did not do the job it was sold to do.

What the orders bar the companies from claiming

The FTC's orders prohibit misrepresentation in four categories, and the definitions matter as much as the categories. The Commission defined Voice Data to include transcripts, not only audio recordings, which closes the obvious gap where a company argues it never handled raw audio.

The orders run for twenty years. That is the part that gives the settlement its weight, because the money involved is modest for a media company and the reporting and compliance obligations are not.

Who pays what

The $930,000 splits unevenly across the three respondents, and PPC Land reported that all of it is already held in escrow by counsel, payable by electronic transfer within eight days of the effective date.

None of the three admitted wrongdoing. That is standard in consent agreements, and it is also the reason a headline about a company being punished overstates what happened. The companies accepted binding conduct restrictions and made payments without conceding the underlying facts.

The procedural detail, for anyone who needs to cite this

Three separate orders were signed by Secretary April J. Tabor under docket numbers C-4838, C-4839 and C-4840, corresponding to file numbers 242-3029, 242-3030 and 242-3033. The Commission approved them 2-0, with Chairman Andrew N. Ferguson and Commissioner Mark R. Meador voting in favour on a Commission that currently has three vacant seats.

The 2-0 vote is worth a second look. With three seats vacant, two commissioners are the sitting Commission, so a unanimous vote here represents everyone currently seated rather than a full five-member body. That is a description of the arithmetic, not a criticism of the outcome, and it is the kind of detail that gets dropped when a case is summarised as the FTC deciding something.

Two public comments were filed on the proposed consent agreements during the comment period. The FTC did not describe their content in the announcement, so what those comments argued is not on the public record as reported. The announcement gives no indication of who filed them or what position they took, and nothing in the report suggests the comments changed the terms that were finalised.

The sequence is worth noting because it is how these cases usually end. A proposed consent agreement is published, a comment window opens, and roughly three months later the Commission finalises it. Nothing about this timeline was unusual, which is itself informative: this was a routine enforcement conclusion, not a dramatic courtroom outcome.

What this case proves, and what it does not

It is easy to read this story as proof that phones are not listening, or as proof that they are. It is neither. The orders concern claims made by three specific companies about one specific product, and the finding is that the product did not have the capability its sellers described.

What the orders cover: these three respondents, those four categories of misrepresentation, for twenty years. What they do not cover: any finding about other advertising vendors, any general conclusion about what consumer devices do or do not capture, and any admission by the respondents that the allegations were true.

The useful takeaway is narrower and more practical than a verdict on whether devices listen. A product was sold on a mechanism nobody could inspect, and the thing that eventually exposed it was not a technical audit. It was the deliverable failing to match the claim, in a way any buyer comparing purchased territory against actual list coverage could in principle have noticed.

What this means for Thai marketers

The belief that your phone is listening is at least as widespread in Thailand as anywhere else, and it shapes how clients react to targeting that works well. This case is a documented example of that belief being packaged and sold, to small businesses, by a media company, with nothing behind it.

The practical habit it argues for is asking vendors what the deliverable actually is, in nouns. When a pitch describes signals, intent detection or proprietary AI, the follow-up question is what arrives: a list, an audience segment, a pixel, a placement. Then where it came from, and how the geography is verified. In the Active Listening case, every one of those questions had an answer that would have contradicted the pitch.

The other habit is measuring against something you own. Campaign performance judged only inside a vendor's own dashboard is judged by the party being paid. A clean analytics setup, with conversions defined by your business rather than by the platform selling you reach, is what lets you tell a targeting product that works from one that is expensive noise.

It also puts a boring virtue in a better light. The major paid social platforms are targeting systems with published controls, documented audience definitions and interfaces where you can see what you selected. That is not a claim that they are perfect. It is a claim that inspectable beats mysterious, and the vendors that resist inspection are the ones this enforcement action came for.

Common questions (FAQ)

Did the FTC find that phones are listening to people?

No. The orders concern the claims these three companies made about Active Listening, and the finding is that the described capability did not exist behind their product. The FTC's action does not establish anything general about what consumer devices capture, and PPC Land's report does not extend the conclusion beyond these respondents.

Did anyone admit to wrongdoing?

No. None of the three respondents admitted wrongdoing. The orders are consent agreements, which means the companies accepted binding restrictions on future claims and made payments without conceding that the allegations were true.

How long do the restrictions last?

Twenty years. The Commission issued final decision and orders on 26 August 2026, converting a $930,000 settlement into obligations running for two decades, and announced the finalisation on 27 August, three months after publishing the proposed consent agreements.

Does the order cover transcripts or only audio recordings?

Both. The FTC defined Voice Data to cover transcripts as well as audio, which means the bar on misrepresenting collection and use of voice data applies to text derived from speech as well as to recordings themselves.

Does this apply to advertising vendors in Thailand?

The FTC is a United States agency and these orders bind three named United States companies, so they carry no direct legal force here. The report said nothing about Thailand. What travels across borders is the pattern rather than the jurisdiction: a targeting claim that cannot be inspected, sold to buyers who had no way to check the deliverable against the pitch.

If a vendor is describing audience data in language you cannot trace to a source, that is worth a conversation before the next contract renewal rather than after it. Relevant Audience is glad to look at what a targeting product is actually delivering against what it promised.

Source: PPC Land.