The US Federal Trade Commission finalized three consent orders on August 27, 2026 over an advertising service marketed as "Active Listening." The agency says Cox Media Group, MindSift, and 1010 Digital Works told small business customers that an AI-powered system could use conversations captured by smart devices to identify nearby consumers who were ready to buy. According to the final complaints, the service did not use voice data and the consumers had not opted into that targeting.
The final orders require payments totaling $930,000 and restrict future claims about marketing services, voice-data collection, consent, and geographic targeting. The case is a practical AI governance signal: an AI label does not reduce the evidence needed for a product claim. It adds questions about what data the system actually uses, what consent exists, and whether a buyer can reproduce the promised result.
What the FTC finalized
The FTC announced proposed settlements in May and finalized them after a public comment period. Cox Media Group must pay $880,000. MindSift and 1010 Digital Works must each pay $25,000. The agency says the money is intended for redress to affected customers.
The orders prohibit the three companies from misrepresenting:
- the qualities or features of advertising and marketing services;
- whether voice data is collected, used, or disclosed;
- whether consumers consented to that collection or use; and
- the geographic targeting capabilities of a service.
These are settlement orders, not a court finding after trial. The companies resolved FTC allegations through consent agreements. That distinction matters when describing the event. The verified fact is that the Commission gave final approval to the orders and the restrictions now apply under those agreements.
Independent reporting by The Desk confirms the final settlement amounts and the agency's account of the service. The underlying factual and legal allegations remain sourced to the FTC complaints and orders.
Why evidence matters before an AI claim goes live
The dispute exposes two separate control failures. The first concerns capability evidence. A vendor claimed that an AI-powered service listened for consumer conversations and converted them into local advertising signals. The FTC says the service actually relied on email lists obtained from data brokers. A buyer approving that product needed evidence connecting the advertised input, the processing method, and the delivered audience.
The second concerns consent. The companies allegedly said consumers had opted into the targeting. The FTC says they had not. Even a technically functioning voice-based service would have created a separate privacy problem if it collected and used conversations without adequate consent.
Teams should therefore review an AI marketing claim as a chain, not as a sentence:
- Data source: data inventory, supplier record, collection method, and permitted use.
- AI capability: system description, test method, version, and reproducible result.
- Consent: exact notice, affirmative choice, scope, timestamp, and withdrawal path.
- Audience outcome: delivery logs, exclusions, geography logic, and error analysis.
- Public wording: approved copy linked to the evidence version used at release.
If one link is missing, the claim should be narrowed or held back.
Controls for vendors and buyers
Marketing teams, procurement owners, privacy officers, and AI governance leads can turn the case into a release gate:
- Inventory the real system. Record every model, data broker, audience file, device signal, integration, and human process that contributes to the service.
- Separate vendor statements from tests. Keep launch copy and sales material, but do not treat them as independent proof. Run a bounded test that can show what inputs produced what outputs.
- Trace consent to the use. A general privacy notice is not automatically evidence of consent for a specific voice or targeting purpose. Record the exact basis and jurisdiction.
- Version the claim. Link each public statement to the product version, dataset, evaluation, and reviewer that supported it.
- Monitor changes. A new data supplier, model, targeting rule, or interface can invalidate prior evidence.
- Preserve challenge records. Keep complaints, exceptions, corrections, and decisions so a team can explain why wording changed.
Maetra's AI compliance evidence checklist offers a structure for owners and current evidence. The AI audit log guide shows how to retain the decision trail behind a consequential claim.
What the orders do not prove
The final orders do not establish a universal rule for every AI advertising product, and they do not certify any alternative targeting method. They address the named respondents and the representations described by the FTC. Applicability to another service depends on its facts, contracts, data flows, claims, and law.
The event also does not show that every use of an AI label is deceptive. It shows why teams need evidence that matches the exact label and promised capability. A claim such as "AI-powered" is still a product representation. If it implies a distinctive data source or performance outcome, the supporting record must cover that implication.
Maetra analysis
The operational lesson is simple: governance starts before publication. A useful control system connects an approved claim to a known system, a lawful data path, current evidence, an accountable owner, and a review trigger. It should also make it possible to withdraw or correct wording when any of those facts change.
For buyers, the same rule belongs in procurement. Do not ask only whether a vendor uses AI. Ask which capability is material, how it was tested, which data makes it work, who consented to that use, and what evidence will remain available after purchase.