Compliance experts flag phrases like "guaranteed," "instant approval," "lowest rate," "no fees," and "first AI [category]" unless they are strictly true and qualified. Unsubstantiated "AI-powered" claims are also risky, so if you claim it, say what that means and provide support.

Yes. In Moffatt v. Air Canada (2024), a tribunal held the airline liable for negligent misrepresentation based on its chatbot's statements, and disclaimers did not shift responsibility. Regulators and courts treat AI outputs as direct corporate statements, so incorrect claims are attributed to your organization.

It is. The SEC issued $400,000 in penalties in March 2024 against two investment advisers for false and misleading AI statements, and the FTC launched Operation AI Comply in September 2024. The FTC's guidance is simple: AI hype does not excuse substantiation.

AI is very good at sounding confident even when it's wrong — it produces plausible but inaccurate statements, exaggerates product benefits, and omits required disclaimers. Because AI produces content faster than manual review can check it, a single flawed output can replicate across thousands of interactions instantly.

Treat your AI like a prolific junior copywriter — require human review for factual claims, rates, fees, and legal qualifiers before anything goes out. It also helps to block risky language at the prompt level, because if the prompt contains risky instructions, the final output is already compromised.

Worqd follows an anti-fabrication policy: until real evidence is approved, we use clearly marked placeholders — we never make up revenue, conversion lifts, logos, or testimonials. Banned words like "guaranteed" and "revolutionary" are kept out of our copy, and human oversight governs every AI system we run.