The Federal Trade Commission's AI enforcement is the most predictable thing in U.S. AI regulation. Read the cases and you have the playbook.
Since launching Operation AI Comply in September 2024, the agency has brought more than a dozen actions under Section 5 of the FTC Act — the same authority that governs every advertising claim made in American commerce. The theory is straightforward: claims about AI capabilities must be substantiated by competent and reliable evidence. The application is new. The standard is not.
Here are the cases that matter and what each one teaches.
DoNotPay — January 2025. The "world's first robot lawyer" settled for $193,000 plus ongoing advertising restrictions. The FTC alleged the AI was inadequately tested, produced ineffective legal documents, and was marketed with capabilities it did not have.
Lesson for brands: "AI-powered" is not a substantiation. Test outputs against the claims. Document the testing. If you can't, kill the claim.
Workado — April 2025. The company's AI Content Detector was advertised as 98% accurate. The FTC alleged actual accuracy was closer to 53%. The model was trained primarily on academic content but marketed as general-purpose. Consent order required Workado to stop making unsubstantiated accuracy claims.
Lesson for brands: If your AI was trained for one use case, you cannot claim it works across all use cases. Training-data scope must match marketing scope.
Growth Cave — January 2026. The FTC alleged the company misrepresented that its "AI software" GrowthBox would automate nearly 100% of online business operations. Settled with permanent restrictions.
Lesson for brands: Automation claims are claims. "AI does it for you" implies performance the system must actually deliver.
Rytr — January 2026 (vacated). The Commission reopened and set aside a 2024 consent order against Rytr, an AI writing tool. The FTC concluded the original complaint failed to allege actual deceptive conduct — the harm was speculative, not concrete.
Lesson for brands: Tools are not illegal because they could be misused. The FTC now requires actual misconduct, not theoretical risk — a narrower standard than the previous administration applied, but not a retreat from enforcement.
Air AI — March 24, 2026. The most aggressive recent action. FTC alleged Air AI bilked entrepreneurs and small businesses out of roughly $19 million through deceptive earnings claims tied to "conversational AI" software marketed as replacing human customer service representatives. Settlement: $18 million monetary judgment (largely suspended), $50,000 in consumer relief, permanent ban on marketing business opportunities.
Lesson for brands: Earnings claims attached to AI capabilities are the highest-risk category in the FTC's portfolio. Refund guarantees that aren't honored compound the violation.
Evolv Technologies — 2024. AI-powered security screening company settled allegations it made false claims about its system's ability to detect threats. The FTC challenged the "AI" framing itself when the underlying detection was substantially less capable than marketed.
Lesson for brands: Calling something "AI-powered" when the AI does limited work is the foundational deception pattern the FTC targets.
The pattern.
Five enforcement theories run through every case:
Inflated performance claims unsupported by testing
Mismatched training and deployment — trained for one thing, sold for another
Automation overstatement — promising the AI does more than it does
Earnings and ROI claims — the highest-penalty category
"AI-washing" — the AI label applied to systems where it does minimal work
What FTC Chairman Andrew Ferguson has signaled.
Ferguson's posture, made explicit at multiple agency forums in late 2025 and early 2026: legitimate AI innovation is protected; the agency will enforce when companies break the law by deceiving consumers. The Rytr walk-back narrowed the theory of speculative harm but did not narrow enforcement of actual deception. Growth Cave and Air AI both proceeded after Rytr was vacated.
What's coming.
The Trump administration's December 2025 executive order directed the FTC to issue a policy statement by March 11, 2026 clarifying how Section 5 applies to AI and when state laws on AI output are preempted. The statement is overdue. When it drops, it becomes the single most important compliance document for brand marketers operating in U.S. markets.
What brands should do now.
Pull every "AI-powered," "AI-driven," "AI-enabled" claim from active marketing. For each one, build a substantiation file — testing methodology, accuracy data, training-data scope, performance benchmarks.
Stop making earnings, ROI, or business-outcome claims tied to AI capabilities unless the numbers are documented and replicable.
Audit influencer and affiliate copy. The FTC's instrumentality theory means a brand can be liable for deceptive claims a partner makes.
Build the substantiation file before the marketing ships — not after a Civil Investigative Demand arrives.
The FTC is telling brands exactly where the line is. The cases are the manual.
