What we found
The Federal Trade Commission announced it will require CMG Media Corporation, doing business as Cox Media Group, along with MindSift LLC and 1010 Digital Works LLC, to pay a total of $930,000 to settle allegations they deceived customers about an "Active Listening" AI-powered marketing service. According to the complaints, the three companies told small businesses that a special algorithm listened in on conversations overheard by consumers' smart devices, in real time, so ads could be targeted to people in a chosen geographic area. The FTC alleged the service did not listen to conversations or use voice data at all, did not accurately place ads in the locations customers asked for, and instead consisted of reselling email lists obtained from other data brokers at a significant markup. The FTC also alleged the companies told potential customers that consumers had "opted in" to the service, when the only consent pointed to was people clicking through mandatory app terms of service. Under the proposed orders, CMG must pay $880,000 and MindSift and 1010 Digital Works must each pay $25,000, money that will be used to provide redress to CMG customers affected by these practices. Before you pay for any ad-targeting service, ask the seller in writing to name the exact data source behind the targeting and to state how the people in it gave consent — and if the answer is only "they agreed to the terms of service," walk away. We do not know from this material how many small businesses bought the service, what individual businesses paid, or how the redress money will reach them. The FTC states it issues an administrative complaint when it has "reason to believe" the law has been or is being violated, and that the proposed consent agreements are subject to 30 days of public comment before the Commission decides whether to make the orders final. LIVEFRAUD ALERTS is independent and not affiliated with the Federal Trade Commission or with any company named here; this alert was assembled only from the single official release listed in our evidence.
- Built on a single official FTC press release naming the firms, the counts, the settlement amounts and the conduct alleged.
- The allegations are set out in administrative complaints and the orders are proposed, not final, so the conduct described is alleged rather than adjudicated.
- No independent reporting or customer accounts were harvested, so the buyer-side experience rests entirely on the agency's summary.
- Reviewed by 3 models, 2 from independent houses.
THE RULE
Small businesses paid a markup for an advertising service that, the FTC alleges, was resold data broker email lists dressed up as real-time eavesdropping — and the consumer "opt-in" behind it was never sought.
What we don’t know
- How many small businesses purchased the Active Listening service, and over what period.
- What individual businesses were charged, and how the markup compared with the underlying data broker list prices.
- How affected CMG customers will be identified and paid redress.
- Whether any consumer voice data was ever collected by anyone in the chain, beyond the FTC's statement that this service used none.
- Whether other firms sold similar "listening" ad products under different names.
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 · MODERATE
The single evidence row supports only the parties, the core deception (voice-listening claim vs. resold email lists), the $930,000 aggregate, and the procedural posture. It does not contain the per-respondent breakdown ($880,000 / $25,000 / $25,000). That allocation is stated as fact and is not carried by the row as written.
DISSENT, PRESERVEDThe headline CLAIM, taken alone, is fully carried by the evidence. My 'overstated' verdict is directed at the elaborated FINDING, which layers in specifics (payment split, consent allegation, geographic-targeting allegation) that a one-row evidence base does not reach. If those details are in the same release, the fix is to expand the evidence rows, not to soften the text.
DISSENT, PRESERVEDI would keep 'alleged' attached to every named company in the lead sentence, not only in the body. Naming three firms as having 'deceived customers' before a final order is the kind of framing that survives corrections poorly.
DISSENT, PRESERVEDReader-risk note: this pattern is unusually persuasive because it exploits a widely held folk belief that phones listen to conversations. That belief does the selling, so the sales pitch needs no evidence of its own. Small-business ad budgets are the exposure, which caps the per-victim loss, but the credulity vector is broad and reusable under other product names.
GROQopenai/gpt-oss-120b · MODERATE
Finding adds numerous specifics—AI‑powered service, real‑time listening algorithm, geographic targeting inaccuracies, significant markup, consent “opt‑in” language, and a detailed payment split—that are not present in the single evidence row.
OPENROUTERopenrouter/free · HIGH
The evidence row does not include details about real-time listening, geographic targeting, significant markup, opt-in claims, or the specific payment breakdown ($880,000 and $25,000 each). These details are present in the finding but are not supported by the provided evidence.
Reviewed by 3 independent models; all judged the finding to go beyond the evidence.
▼ Protocol & challenge record
Objection (high): The directive_options list contains "Send this to any christopher you know." and "Forward this to the christophers in your life." The only "Christopher" in the harvested row is Christopher Mufarrige, Director of the FTC's Bureau of Consumer Protection — the quoted official. A name-extraction routine has evidently mistaken an FTC official's given name for a victim/target cohort. Publishing either option would tell readers to forward a fraud alert to people named Christopher, a fabricated targeting claim derived from a government spokesman. These options must be deleted, not merely left unselected, and the extraction bug logged.
Resolved: Delete both "christopher" directive options outright; do not leave them selectable. File the underlying defect: given-name extraction is pulling from the quoted-official field. Replace with the evidence-supported "Forward this to a small business that bought local ad targeting."
Objection (high): callout_options includes "ATTENTION: VETERANS." Nothing in the harvested row mentions veterans, military affiliation, or any demographic cohort. This is exactly the class of unsourced claim-about-people the pipeline says it dropped elsewhere (targeting_dropped, §11 Rule 2). Its presence as a selectable option is a live risk of publishing an invented victim group.
Resolved: Delete "ATTENTION: VETERANS" from callout_options. Retain "ATTENTION: SMALL BUSINESSES BUYING ADS" or "ATTENTION: ANYONE SOLD AI AD TARGETING," both of which map to the row's "small businesses" language.
Objection (medium): Internal inconsistency in targeting: the pipeline dropped a "buyers" callout and an "owner" directive as "NOT IN EVIDENCE," yet the harvested row explicitly and repeatedly identifies the affected group — "allow small businesses to advertise," "misled small businesses about the capabilities." Small-business ad buyers ARE in evidence. The drop rationale is factually wrong against this row, and the finding itself ("told small businesses") plus the advice sentence both address that group anyway. Either the drop is a false positive or the finding/advice are inconsistent with it.
Resolved: Reverse the "buyers" drop and cite row 7c5e7c85 ("allow small businesses to advertise"; "misled small businesses"). Select the small-business callout. Keep the "owner" directive dropped only if no source describes owners specifically.
Objection (medium): The claim line states the allegation as established fact in its second clause: "...when the service actually resold email lists bought from data brokers." The word "actually" converts an unadjudicated administrative-complaint allegation into a finding. The orders are proposed, the complaints are "reason to believe" pleadings, and the release records no adjudication. The claim, which may be surfaced standalone without the finding's later hedge, needs "the FTC alleges" inside it. Also "bought from data brokers" overstates the source's "obtained from other data brokers."
Resolved: Rewrite the claim as: "The FTC has proposed settlements requiring Cox Media Group and two marketing firms to pay $930,000 over allegations they sold small businesses an 'Active Listening' ad service they falsely claimed used conversations overheard by smart devices; the FTC alleges the service instead resold email lists obtained from data brokers."
Objection (medium): The claim/headline "The FTC will require ... to pay $930,000 to settle" presents a completed obligation. The source's own body makes clear the consent orders are proposed and the Commission will only "decide whether to make the proposed consent orders final" after a 30-day comment period. The FTC's headline uses the same shorthand, but a consumer-facing alert repeating it without the word "proposed" in the claim overstates finality.
Resolved: Insert "proposed" into the claim and keep the existing procedural sentence in the finding, moving it earlier so the 30-day comment caveat is not the penultimate line.
Objection (medium): Date currency: the sole source is dated 2026-05-21 and the finding asserts in present tense that the agreements "are subject to 30 days of public comment." That window runs from Federal Register publication, which was still pending at press time. If the alert is being published more than roughly six to eight weeks after 2026-05-21, the comment period has likely closed and the Commission may already have finalized the orders — making the finding's procedural posture stale and the "proposed" framing potentially wrong in the other direction. The draft records no check of the docket or Federal Register for a subsequent final-order announcement.
Not resolved — preserved on the record.
Objection (medium): The risk_line hedges the middle clause ("the FTC alleges") but ends unhedged: "and the consumer 'opt-in' behind it was never sought." That is an allegation from the complaints stated as fact, in the most quotable line of the item. Fix: "...and, the FTC alleges, consumers' consent was never sought."
Resolved: Amend risk_line to place the allegation marker at the head and keep it governing both clauses: "The FTC alleges small businesses paid a markup for data-broker email lists dressed up as real-time eavesdropping, and that the consumer 'opt-in' behind it was never sought."
Objection (low): The advice sentence carries no row_ids and ends with a hard prescription — "walk away." The harvest supports the observation that click-through terms of service are not opt-in consent (the FTC says so explicitly), but it supports nothing about asking sellers in writing to name data sources, and it does not support a blanket instruction to abandon a purchase. This is generic editorial counsel presented in the voice of a sourced alert. At minimum it should be marked as the outlet's own guidance rather than sitting adjacent to sourced sentences.
Not resolved — preserved on the record.
Objection (low): watch_icons includes "phone." The harvested row describes voice data from smart devices and resold email lists — no telephone or SMS vector at all. "email" and "person" are defensible; "phone" implies a contact channel the evidence does not establish.
Resolved: Drop the "phone" icon; retain "person" and "email," or substitute a voice/device icon if the set supports one.
Objection (low): Material omissions that change the picture for a reader: (a) the second count against MindSift and 1010 Digital Works for supplying CMG the "means and instrumentalities" to deceive — which is why the two small firms are on the hook at all; (b) the conduct bans in the proposed orders (no misrepresenting service features, voice-data collection/consent, or geo-targeting), which are the durable remedy; (c) the Commission voted 2-0, i.e. a two-member Commission. None are errors, but a $930k figure without the injunctive terms undersells the outcome.
Not resolved — preserved on the record.
Objection (low): The evidence row's limitation says "the release records no admission of wrongdoing." The release is simply silent on admissions; it neither records nor denies one. Stating the absence as if it were a documented feature of the settlement is a small inference beyond the text. Rewrite as "the release does not state whether the companies admitted wrongdoing."
Resolved: Reword the evidence limitation to "the release does not state whether the companies admitted wrongdoing, and no final order has issued."
Objection (low): Confidence is "high" while two of the three confidence_reasons are reasons for caution (unadjudicated allegations; single source, no buyer-side corroboration). The rating is defensible for the narrow proposition "the FTC announced this action," but it is not high for "the service worked this way." The confidence object should say which proposition it is rating.
Resolved: Split the confidence statement: high confidence that the FTC announced the action and the figures are as stated; low-to-moderate confidence about the underlying conduct, which rests on unadjudicated allegations from a single agency summary.
Preserved dissent
ON THE RECORDThe two 'christopher' directive options are not a stylistic quibble. They are a name lifted from the FTC's own Bureau of Consumer Protection director, quoted in the press release, and turned into a purported victim cohort. If either had shipped, the alert would have instructed readers to warn people who share a first name with the government official announcing the case. I want it on the record that this is an extraction defect in the pipeline, not a one-off, and that suppressing the two options without fixing the extractor leaves the same failure available on the next official quote.
ON THE RECORDI also dissent from the 'buyers not in evidence' drop. The harvested row names small businesses as the marketed-to and misled group four separate times. Applying §11 Rule 2 to strike that callout while the finding freely says 'told small businesses' is not caution, it is inconsistency, and it strips the alert of the one piece of targeting the source actually supports.
ON THE RECORDOn confidence: 'high' is fine for what the FTC announced and wrong for what the companies did. Every operative fact in this item is an allegation in a proposed administrative complaint decided by a two-member Commission with no adjudication. The draft's own confidence_reasons say so and then rate it high anyway. I would not sign the current rating without the split.
The sources
Official sourceFTC 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 Service2026-05-21
The FTC alleged that CMG (dba Cox Media Group), MindSift LLC and 1010 Digital Works LLC marketed an "Active Listening" service as targeting ads from conversations captured by smart devices when it used no voice data and instead resold email lists, and will require the firms to pay $930,000 in settlement.
Authority: official. Retrieved 2026-08-20.
Limitation: These are allegations in proposed administrative complaints subject to 30 days of public comment; the release records no admission of wrongdoing and no final order.
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 to Require Cox Media Group, Two Other Firms to Pay Nearly $1 Million to Sett".
- ✓ All 6 material sentence(s) map to FTC.
- ✗ anthropic returned "overstated"; groq returned "overstated"; openrouter returned "overstated" — 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 2026-08-20.
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