On May 21, 2026, the U.S. Federal Trade Commission announced consent agreements with Cox Media Group (CMG Media Corporation), MindSift, and 1010 Digital Works: the three firms will pay a combined $930,000 to settle charges that they deceived customers with an AI-powered advertising service marketed as “Active Listening.” It is the latest entry in the FTC’s growing “AI washing” docket, and the facts are unusually ironic — the AI that was sold as eavesdropping on consumers’ conversations never listened to anyone at all.
The case deserves a close read from any AI product or marketing team. The FTC’s theory of harm is not that the listening was invasive; it is that the “AI capability” was fictional from the start.
What “Active Listening” Claimed to Do
The story begins with a leaked pitch deck. In 2024, a Cox Media Group sales presentation surfaced (it now lives on DocumentCloud) telling advertisers that smart devices capture real-time intent signals from consumers’ conversations. Advertisers could supposedly pair this “voice data” with behavioral data to target in-market consumers in specific geographic areas — and consumers had consented. The service had a name built for a slide: Active Listening.
Security researcher Simon Willison called it out at the time, arguing it looked like an ordinary ad-targeting product dressed up in spy-thriller language. The FTC’s investigation proved him right.
What the FTC Actually Found
According to the complaints, the service never used voice data and never eavesdropped on a single conversation. What the companies actually did was buy email lists from other data brokers and resell them to clients at a significant markup. The service even failed to place ads in the geographic regions clients requested.
The “consumers opted in” defense fared no better. The FTC rejected the argument that consumers agreed to anything by clicking through mandatory app terms of service — forced acceptance of a ToS is not opt-in consent for a service this invasive. The agency added a sharper point: had the service actually worked as advertised, collecting voice data without adequate consent would itself have violated Section 5 of the FTC Act. Either way, the road was illegal.
The Settlement Terms and Penalties
The money breaks down as follows: Cox Media Group pays $880,000, while MindSift and 1010 Digital Works pay $25,000 each, for a total of $930,000 earmarked for redress to affected CMG customers. The consent orders bar all three companies from misrepresenting three categories of facts:
- The qualities or features of their advertising and marketing services
- The collection and use of voice data, and whether consumers consented to it
- The geographic targeting capabilities of their advertising services
Future violations of a final order can trigger civil penalties of up to $53,088 each. MindSift and 1010 Digital Works face a second charge for supplying CMG with the “means and instrumentalities” of deception. The Commission approved the complaints and settlements on a 2-0 vote, and the agreements enter a 30-day public comment period after Federal Register publication. Christopher Mufarrige, Director of the FTC’s Bureau of Consumer Protection, kept the statement blunt: “It is a basic rule of business that you need to be honest with your customers, and these companies failed to do that.”
Why “AI-Powered” Claims Became the Enforcement Hook
The enforcement logic here matters for every company that slaps “AI-powered” onto a landing page. The FTC did not need to prove an AI system misfired — the injury is that the capability was fake. Wrapping a data-broker list resale in an “AI listening engine” narrative is deceptive under Section 5 in exactly the same way as inflating performance numbers.
The second lesson concerns consent architecture. Treating mandatory terms of service as “consent” did not survive contact with the FTC. Any product planning to personalize off voice, behavioral, or biometric data needs a consent flow that holds up to regulatory scrutiny — not a disclosure buried in an FAQ page.
Three Lessons for Marketing and Product Teams
First, capability claims must match what you actually ship. The fraud here was not the listening; it was that no listening existed, and every sentence in that pitch deck that oversold the system is now an exhibit. Second, “consumers consented” requires a real mechanism: popups and forced ToS acceptance are not consent. Third, do not monetize public paranoia. The widespread belief that phones are always listening made the “we listen” pitch feel plausible — and it also guarantees that the reputational and legal bill is brutal when the claim collapses.
Willison notes a useful side effect: the FTC’s findings are an official rebuttal of the “microphone ads conspiracy.” What actually happened was not surveillance but a plain scam. For developers and product builders, this is a textbook case that the regulatory cost of AI storytelling is now as real as the technical risk of AI itself.
Sources
- FTC to Require Cox Media Group, Two Other Firms to Pay Nearly $1 Million to Settle Charges They Deceived Customers — FTC
- FTC active listening settlement — Simon Willison
- CMG pitch deck on voice data advertising (Active Listening) — DocumentCloud
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.