DoNotPay, Inc..; Analysis of Proposed Consent Order To Aid Public Comment
The consent agreement in this matter settles alleged violations of federal law prohibiting unfair or deceptive acts or practices. The attached Analysis of Proposed Consent Order to Aid Public Comment describes both the allegations in the complaint and the terms of the consent order--embodied in the consent agreement--that would settle these allegations.
What this rule actually says
The FTC settled a case against DoNotPay for making false claims about what their AI could do—specifically, they claimed their chatbot could beat traffic tickets and sue robocallers, but it mostly just generated template letters that users still had to file themselves. The consent order requires the company to stop making exaggerated claims about AI capabilities and to have evidence backing up any performance claims they make going forward.
Who it applies to
- If you claim your AI can do something specific (win cases, get refunds, diagnose conditions, etc.), this applies to you.
- If you're in the US, this applies. The FTC enforces this nationwide.
- Use cases that trigger it: Any AI product where you make performance claims—medical diagnosis accuracy, hiring prediction rates, customer issue resolution rates, legal outcome success rates.
- What's in scope: Claims about what your product accomplishes, success rates, capabilities, and results.
- What's out of scope: General descriptions of features (e.g., "uses AI to summarize notes") without specific performance claims.
What founders need to do
- Audit your marketing claims (2-3 days). Review your website, pitch deck, demo videos, and emails. Flag anything that claims specific outcomes: "resolves 80% of support tickets," "predicts hiring fit," "catches 95% of errors." If you can't back it up with real data, remove it.
- Document your evidence (1-2 weeks). For any performance claim you keep, gather the testing or real-world data that supports it. This doesn't need to be a published study—internal benchmarks, pilot results, or customer metrics count, but you need *something* written down.
- Rewrite claims conservatively (2-3 days). Change "Our AI diagnoses X" to "Our AI helps clinicians by flagging potential X" or "This tool assists with X tasks." Shift from guarantees to assistance framing.
- Set a quarterly review (ongoing, ~2 hours per quarter). As your product evolves, make sure new claims stay defensible. This is a living document, not a one-time fix.
- If you're already claiming results, consult a lawyer (1 week). If your product is live and making specific performance claims, have a lawyer do a quick review—the cost is worth the liability reduction.
Bottom line
Act now if you're making specific performance claims about your AI; monitor if you're still pre-launch or only describing features generically.