Every week another tool shows up in your inbox promising to be "AI-powered." Some of them are. Some of them aren't, and regulators have started proving it in public. Over the past two years the Federal Trade Commission has brought a string of cases against companies that sold AI hype instead of AI, and two of the most recent ones read like a warning label for anyone shopping for AI tools right now.
The company that promised an AI that would replace your staff
The FTC's case against Air AI is the one small-business owners should pay closest attention to, because Air AI marketed directly to entrepreneurs and small businesses rather than big enterprise buyers. According to the FTC's complaint, the company pitched a voice-AI product as a drop-in replacement for human customer-service staff and paired that pitch with promises that buyers could earn back their investment, and then some, in a matter of weeks. The FTC's settlement bans Air AI and its owners from marketing or selling any "business opportunity" going forward and from making unsubstantiated earnings claims of any kind.
Notice what actually got Air AI in trouble. It wasn't that the underlying AI didn't exist — it was the layer on top of it: guaranteed outcomes, inflated performance claims, and revenue promises no vendor can actually back up for a business they've never seen operate. That's a pattern worth watching for anywhere, not just in one case.
The "AI" that was just a spreadsheet of purchased emails
The second case is almost funnier, in a grim way. Cox Media Group and two partner marketing firms told advertisers they had built an "Active Listening" tool that used AI to analyze conversations picked up by smart devices — phones, TVs, voice assistants — and turn them into hyper-targeted local ads. The FTC's investigation found no such listening technology existed. What the companies were actually selling was ordinary email lists bought from data brokers, repackaged and marked up under an AI label that made the targeting sound futuristic and precise.
Both cases are part of what the FTC calls Operation AI Comply, an ongoing enforcement push against what regulators now call "AI-washing" — dressing up an ordinary product in AI language to make it sound more capable, more automated, or more valuable than it actually is. This isn't a one-off crackdown; law firms tracking the agency's activity describe it as a sustained enforcement focus that has continued across multiple cases, not a single headline that will fade.
What this means when you're the one buying
You don't need a legal team to protect yourself from a vendor that oversells. You need better questions before you sign anything. A few that work well in practice:
- Ask for a live demo on your own data or your own scenario, not a polished recorded walkthrough. A tool that only performs well in a canned demo is telling you something.
- Ask what model or technology actually powers the product. A vendor that can't or won't answer — "it's proprietary" is not an answer — is a red flag. Legitimate vendors will tell you, in plain terms, whether they're built on a known model (like GPT, Gemini, or Claude) or a custom system, and what it does and doesn't do well.
- Be specific about outcomes, in writing. "Saves time" and "increases sales" are marketing. Ask what specifically improves, for what kind of business, and ask if they'll put a specific, testable claim in the contract.
- Treat guaranteed earnings or ROI numbers as a warning sign, not a selling point. No one can guarantee what a tool will do for your specific business before they've seen it run there. That's exactly the claim that got Air AI banned from making it.
- Ask what happens if it's wrong. Every real AI tool makes mistakes sometimes. If a vendor can't describe how errors get caught — a human review step, a confidence threshold, an escalation path — the product probably isn't as automated as the pitch suggests, or it is, and nobody's watching it.
- Start small. A short paid pilot against a real task in your business will tell you more in two weeks than any sales deck will tell you in two hours.
The point isn't to be afraid of AI tools
Plenty of the AI tools on the market do exactly what they say. The point of watching these cases isn't to get cynical about AI — it's to get specific. The businesses that got burned by Air AI and by the "Active Listening" pitch weren't burned by AI; they were burned by claims nobody checked before money changed hands. A vendor with a real product will answer specific questions about how it works, what it costs, and what happens when it fails. One that won't is telling you exactly what you need to know before you buy.