What we found
The Federal Trade Commission finalized orders requiring CMG Media Corporation, which does business as Cox Media Group, and two marketing firms it worked with, MindSift LLC and 1010 Digital Works LLC, to pay a total of $930,000 to settle allegations they deceived customers about an AI-powered marketing service. According to the FTC's complaints, first announced in May, the companies told customers they used a special algorithm to listen in on and detect pertinent conversations from smart devices in order to target ads to consumers within a specific geographic region. The FTC states that, contrary to those claims, the marketing service was not based on voice data and consumers had not opted into it, and that if the service had worked as advertised, collecting and using voice data without adequate consent would itself violate the FTC Act. Under the orders, CMG must pay $880,000 while MindSift and 1010 Digital Works must each pay $25,000, money the FTC says will be used to provide redress to CMG customers impacted by these practices. Each defendant is also prohibited from making misrepresentations about the features of its advertising services, about the collection and use of voice data and consumer consent to it, and about geographic targeting capabilities, and the Commission voted 2-0 to give final approval after receiving two comments. If a marketing pitch tells you it can aim ads at what a phone or smart speaker overheard, ask the seller to put the actual data source in writing before you sign or pay anything. The FTC release does not say how affected customers will be identified or contacted about redress payments. Our harvested material describes an enforcement action against sellers of an advertising service and does not describe anyone impersonating these companies or contacting consumers directly. LIVEFRAUD ALERTS is independent and not affiliated with the Federal Trade Commission or with any company named here; the companies are named only because the official record names them, and this alert rests on the single agency release listed above.
- The underlying facts come directly from a primary official source: an FTC press release announcing finalized consent orders, with named respondents, dollar amounts and a recorded 2-0 Commission vote.
- Only one harvested row is available, so nothing here is corroborated by a second independent source.
- The harm described falls on business customers who bought the advertising service, not on consumers contacted by a scammer, so the consumer-facing risk is inferred from the pitch itself rather than described by the source.
- Reviewed by 3 models, 2 from independent houses.
THE RULE
The money lost here was paid by advertising customers for a targeting capability the FTC says did not work as sold — treat any "your device is listening" ad pitch as a claim to be substantiated before payment.
What we don’t know
- How many customers bought the advertising service, and how much each paid.
- How impacted CMG customers will be identified, notified or paid redress.
- Whether any consumer voice data was ever collected by any party at any point.
- What the service actually used to target ads, if not voice data.
- Whether other firms are marketing similar "active listening" targeting claims.
The bench — who voted
3 INDEPENDENT AI MODELS REVIEWED THIS. 2 OF 3 — ONE MODEL DISSENTED. READ WHY.
The card names a count. Here are the seats behind it, with what each one said.
ANTHROPICclaude-opus-5 · LOW
The single evidence row says only 'Cox Media Group and two marketing firms.' The finding names CMG Media Corporation, MindSift LLC and 1010 Digital Works LLC as the parties alleged to have deceived customers. Naming specific legal entities as respondents in a deception action is exactly the kind of assertion that must sit on the record text, and here it does not.
DISSENT, PRESERVEDThis is filed as a consumer fraud alert but the harvested material describes a business-to-business enforcement action against sellers of an advertising product. The people deceived were advertising customers, not the device owners. Publishing it under a consumer-alert banner risks readers taking away the opposite of the FTC's finding — that their phones and speakers were in fact listening — when the sourced line says the service was not based on voice data at all. The protective advice offered is addressed to purchasers of ad services and is of little use to a general reader.
DISSENT, PRESERVEDThe aggregate dollar figure and the fact of finalization hold cleanly. If the panel elects to publish, I would strip the unsourced entity names, the payment split, the vote tally, the comment count and the injunctive terms, and run only what the single row carries.
DISSENT, PRESERVEDThe stated confidence of 'moderate' is not reconcilable with a finding built on one press release and padded with a dozen particulars that release is not shown to contain.
GROQopenai/gpt-oss-120b · MODERATE
Finding specifies payment breakdown ($880k, $25k each) not present in the evidence row, which only states total $930k.
OPENROUTERopenrouter/free · MODERATE
The finding's specific payment breakdown ($880,000 for CMG, $25,000 each for the other defendants) is not present in the single evidence row, though it is stated in the full FTC release. This is minor — the evidence row confirms the $930,000 total figure and the nature of the allegations.
Reviewed by 3 independent models: 1 found it carried by the evidence, 2 did not.
▼ Protocol & challenge record
Objection (high): risk_line overclaims twice. (a) "The money lost here was paid by advertising customers" — the source never says any customer lost money. It says respondents will pay sums that "will be used to provide redress to CMG customers impacted by these practices." Redress funding is not a finding of quantified customer loss, and the release does not say how many customers bought the service or what they paid (the draft's own unknowns concede this). (b) "a targeting capability the FTC says did not work as sold" stretches the source. The FTC says the service "wasn't based on voice data" and that consumers hadn't opted in — i.e., the represented mechanism was false. It does not say the ads failed to run, failed to target geographically, or that the service delivered nothing. Rewrite to: "the FTC says the service was not based on voice data as advertised."
Not resolved — preserved on the record.
Objection (medium): Settlement framing is missing the standard qualifier. These are consent orders resolving allegations; the source paragraph describing the $880k/$25k/$25k payments explicitly says "Under the proposed orders settling the FTC's allegations." The draft's body renders this as "Under the orders" and "contrary to those claims, the marketing service was not based on voice data" — the latter is attributed to the FTC in the lead-in, which is acceptable, but nowhere does the alert note that the companies settled without any adjudication or admission of liability. For an alert that names three specific companies, that omission is material.
Not resolved — preserved on the record.
Objection (medium): watch_icons ["phone","person","link"] imply a contact vector that this record does not contain. There is no phone outreach, no impersonation, no malicious link in this enforcement action — the draft's own limitation sentence says exactly that. Iconography signalling "phone/person/link" contradicts the limitation and risks readers scanning the alert as a consumer contact scam. Either drop the icons or use ones tied to advertising/deceptive-claims.
Resolved: Partially self-mitigated in text: the limitation sentence explicitly states the material "does not describe anyone impersonating these companies or contacting consumers directly." The icons still contradict that sentence and should be changed, but a careful reader is warned.
Objection (medium): Date ambiguity carried through unresolved. The release says complaints were "first announced in May" with no year; the finalization is dated 2026-08-27. The draft repeats "first announced in May" without flagging that the year is not stated in the source. A reader will default to May 2026, which may be wrong. Either state "in May (year not specified in the release)" or omit.
Resolved: Draft faithfully mirrors the source's own vagueness rather than inventing a year; the defect is inherited, not manufactured. Fixable with a four-word parenthetical.
Objection (medium): Unexplained asymmetry not surfaced. CMG pays $880,000 (94.6% of the total) and redress goes only to "CMG customers," yet MindSift and 1010 Digital Works are described flatly as "two marketing firms it worked with" facing identical conduct allegations. The source does not explain the difference in role or exposure, and the draft neither notes nor lists it as an unknown. Naming two small firms alongside a large media company without that caveat overstates their comparable culpability.
Not resolved — preserved on the record.
Objection (low): Option lists contain choices unsupported by any harvested content. "ATTENTION: VETERANS" and "ATTENTION: EVERYONE" have zero basis in a B2B advertising-deception record, and three of four directive_options ("Send this to any consumer you know," "Forward this to the consumers in your life," "Send this to any customer you know") point at consumers who are not the injured party here. The selected callout and directive are correct; the option pool is not, and a wrong pick would produce a materially misleading alert.
Resolved: The selected audience_callout ("BUSINESSES BUYING AD SERVICES") and share_directive correctly identify the injured party as ad buyers, so the shipped alert is not mistargeted; the objection is to the option pool, not the output.
Objection (low): Minor source-to-claim drift in the prohibition sentence: the order bars misrepresentation about the "qualities or features of its advertising or marketing services"; the draft narrows this to "features of its advertising services," dropping both "qualities" and "marketing." Small, but it understates the scope of the injunctive relief.
Resolved: Meaning is preserved in substance; the narrowing is a compression artifact, not a factual error.
Objection (low): The advice sentence ("ask the seller to put the actual data source in writing before you sign or pay anything") is unsourced and marginally actionable — a seller willing to fabricate an active-listening algorithm in a pitch deck is not deterred by being asked to repeat it in writing. It is not wrong, but it is presented as the operative protective step and it isn't one.
Not resolved — preserved on the record.
Objection (low): confidence_reasons conflate corroboration count with reliability. "Only one harvested row is available, so nothing here is corroborated" is true but weak grounds for downgrading: a primary FTC press release announcing its own final consent orders is self-authenticating for the existence, parties, amounts and vote. The genuine reason for restraint is the inferential leap to consumer-facing risk, which is reason three. As written, the confidence rationale is partly padded.
Not resolved — preserved on the record.
Preserved dissent
ON THE RECORDI do not accept the risk_line as written. "The money lost here was paid by advertising customers for a targeting capability the FTC says did not work as sold" asserts two things the single harvested row does not support: that customers lost money (the release establishes a redress fund, not a quantified loss) and that the service "did not work as sold" (the release says only that it was not based on voice data and that consumers had not opted in). The draft's own unknowns list concedes it does not know how many customers bought the service, what they paid, or what the service actually used. A risk line should not assert what the unknowns section disclaims.
ON THE RECORDThe icon set [phone, person, link] should not ship. This is a business-to-business deceptive-advertising enforcement action with no consumer contact vector, and the alert's own limitation paragraph says so. Attaching contact-scam iconography to it is the single most likely way a reader misreads this alert.
ON THE RECORDI would flag, and the desk has not, that all three named companies settled without admission of liability. Publishing three company names under a fraud-alert banner without that one clause is an avoidable fairness gap, and it costs eight words to close.
The sources
Official sourceFTC Finalizes Orders with Cox Media Group, Two Other Firms Settling Charges They Deceived Customers About “Active Listening” AI-Powered Marketing Service2026-08-27
The FTC finalized orders requiring Cox Media Group and two marketing firms to pay $930,000 over allegations they falsely claimed an AI service could target ads using conversations captured from consumers' smart devices, when the service was not based on voice data and consumers had not opted in.
Authority: official. Retrieved 2026-09-01.
Limitation: A single agency press release; it does not name affected customers, does not quantify how many customers bought the service, and does not state that any consumer voice data was in fact collected.
Open the original source →Other checks
Published under standing founder pass (A9) — every claim source-mapped by the machine.
▼ What the machine checked
- ✓ Not a community submission.
- ✗ Draws on an FTC enforcement release, which names a defendant: "FTC Finalizes Orders with Cox Media Group, Two Other Firms Settling Charges They".
- ✓ All 5 material sentence(s) map to FTC.
- ✗ anthropic returned "overstated"; groq returned "overstated"; openrouter raised 1 objection(s) — published on the receipt, not blocking (A9 amendment).
- ✓ No audience band is set.
No human affirmed these. They were verified by the classifier described in Amendment A9, on 2026-09-01.
Something wrong here? Tell us and we'll correct it — corrections are published, not quietly edited.
Phishy? Send it → sharelivefraud.com/squire-it
Not affiliated with any government agency, credit bureau, bank, platform, or law-enforcement agency. Informational only — not legal or financial advice.
Naming a source is not an endorsement, and being named here is not an accusation against any company.
Powered by SquireIt™
Verify this receipt at squireit.com