Summary: On May 21, 2026, the FTC announced consent agreements under which Cox Media Group, MindSift, and 1010 Digital Works will pay $930,000 over deceptive marketing of an AI ad service called Active Listening. The investigation found the service never used voice data or listened to conversations; it resold purchased email lists. The orders bar false claims about capabilities, voice-data collection and consent, and geographic targeting.
On May 21, 2026, the U.S. Federal Trade Commission announced consent agreements with three companies — Cox Media Group (CMG Media Corporation), MindSift, and 1010 Digital Works — that will together pay $930,000 to resolve charges they deceived customers with an AI advertising service sold under the name “Active Listening.” The case is the newest page in the FTC’s “AI washing” enforcement ledger, and its facts come with a sharp dose of irony: the AI pitched as eavesdropping on consumers’ conversations never actually listened to a single person.
Any AI product or marketing team should study this one closely. What the FTC is penalizing is not that the listening would have been invasive; it is that the “AI capability” was fiction from day one.
Where “Active Listening” Came From: A Leaked Pitch Deck
The story opens with a Cox Media Group sales presentation that leaked in 2024 (the document now sits on DocumentCloud). Advertisers reading it were told that smart devices pull real-time intent signals out of consumers’ conversations. Blend that “voice data” with behavioral data, the deck claimed, and advertisers could reach in-market consumers across chosen geographic areas — with consumers, it said, already consenting. The service carried a name made for the slide: Active Listening.
Security researcher Simon Willison raised doubts at the time, arguing the whole thing looked like ordinary ad targeting dressed up in lurid surveillance language. The FTC’s investigation bore him out.
What the Investigation Turned Up: No Listening, Just Resold Lists
Per the complaints, the service never touched voice data and never eavesdropped on a single conversation. The actual business was mundane: the companies bought email lists from data brokers and resold those lists to clients at a hefty markup. Even the geographic placement clients asked for frequently failed to materialize.
The “consumers opted in” defense held up no better. Forced clicks through an app’s mandatory terms of service, the FTC concluded, do not amount to opt-in consent for a service this intrusive. The agency then turned the screws: had the service worked exactly as advertised, collecting voice data without adequate consent would by itself have broken Section 5 of the FTC Act. Either path led to illegality.
Settlement Money and the Barred Claims
Here is how the money splits: Cox Media Group owes $880,000, MindSift and 1010 Digital Works owe $25,000 apiece, and the combined $930,000 is reserved for redress to affected CMG customers. Under the consent orders, none of the three companies may misrepresent facts in three categories:
- What their advertising and marketing services are actually capable of
- How voice data gets collected and used, and whether consumers consented
- How far their advertising services can target by geography
A future violation of a final order can bring civil penalties of up to $53,088 per violation. MindSift and 1010 Digital Works also carry a second charge: supplying CMG with the “means and instrumentalities” of deception. The Commission signed off on the complaints and settlements by a 2-0 vote, and once the agreements publish in the Federal Register, a 30-day public comment period follows. Christopher Mufarrige, the FTC’s Bureau of Consumer Protection Director, kept his comment blunt: “It is a basic rule of business that you need to be honest with your customers, and these companies failed to do that.”
How Fake AI Claims Became the Enforcement Hook
The enforcement logic is one every company sprinkling “AI-powered” over its landing pages should absorb: the FTC never had to show an AI system malfunctioning, because the injury is the capability being fake. Wrapping list resale in an “AI listening engine” narrative is deceptive under Section 5 in precisely the way inflated performance numbers are.
The second strand is consent design. Mandatory terms of service posing as “consent” did not survive contact with the FTC. Any product that plans to personalize using voice, behavioral, or biometric data needs a consent flow that holds up under regulatory examination — not a disclosure tucked away in an FAQ.
Three Takeaways for Marketing and Product Teams
One: claims must equal what you ship. The fraud was never the listening — it was that no listening existed, so every line in that pitch deck that oversold the system now stands as an exhibit. Two: “consumers consented” needs a working mechanism behind it; popups and forced ToS acceptance are not consent. Three: don’t do business on top of public paranoia. The common belief that phones are always listening made the “we listen” pitch sound believable — and the same belief guarantees a brutal reputational and legal bill when the claim collapses.
Willison adds a useful footnote: the FTC’s findings amount to an official debunking of the “microphone ads conspiracy.” What really took place was not surveillance but plain fraud. The textbook message for developers and product builders is that the regulatory cost of AI storytelling now rivals the technical risk of AI itself.