FTC Act Section 5 (15 U.S.C. § 45); Operation AI Comply (2024)
For marketing and product teams describing AI features working out how the FTC treats claims about what AI can do. This page covers what you need to know and ways to avoid running into issues.
What we've seen reported
AI washing, a name for overstated or unsupported claims about what AI products can do, is treated by the FTC as deceptive advertising under Section 5 of the FTC Act, which it applied in its 2024 Operation AI Comply sweep.1,2
Last updated October 2026 · General information only, not legal advice. Sources.
What the rules cover
These are the kinds of AI claims the FTC has described in its guidance and Operation AI Comply cases under Section 5 of the FTC Act, and how they tend to show up in marketing.2,3
| Phrase or term | In practice |
|---|---|
| "AI-powered", "built on AI" | FTC staff guidance warns against saying a product uses AI when it doesn't. |
| "Smarter than a human", "replaces your lawyer" | Claims that AI outperforms people or non-AI products need proof, the FTC says. |
| "98% accurate" | An accuracy figure without competent and reliable evidence behind it. |
| "Earn passive income with AI" | Earnings claims tied to AI tools, the focus of several Operation AI Comply cases. |
| "Our AI does it all" | Exaggerating what an AI product can do, or promising results it can't deliver. |
| "AI-driven investment process" | The SEC has settled charges over advisers' false statements about using AI. |
The FTC has already finalized orders under Section 5 over claims about AI products.
An online service marketed as an AI lawyer agreed to an FTC order barring claims it could stand in for a human lawyer, with monetary relief.
A company selling an AI content detection tool agreed to an FTC order over accuracy claims the FTC said were unsupported.
Where it goes wrong
One approved claim ends up in many channels, markets and languages. Along the way it gets shortened, translated and reworded, and some versions end up using the very phrases you are trying to avoid.
"Heldan Assist drafts replies in your team's approved style. Your team reviews every reply."
How Arbitr helps
Manually review everything? Or do what modern teams do and use AI to check that what you publish matches what you approved.
FAQ
Wherever a product gets described quickly. The FTC's 2023 guidance on AI claims asks marketers whether they're exaggerating what an AI product can do,3 and we usually find the boldest versions in:
Agree the approved description of each AI feature with product before launch, including what it doesn't do. Then check every new version against it, because accuracy figures and "replaces" claims tend to appear in ads and decks months later. Our content checks page shows how teams check new content against approved wording.
Keep the approved wording, the test results behind any accuracy or performance figure, the date of the testing, who approved it and where it's published. The FTC's guidance expects proof for claims that AI does something better than a non-AI product,3 so test dates matter when the product changes.
How it applies to a specific claim is a question for your legal adviser. Where that claim appears across your markets and languages is a question Arbitr answers.
Product marketing usually writes these claims, but sales, partnerships, investor relations and agencies repeat them. A short briefing using your own AI features as examples helps. Rules on labeling AI-generated content, such as the California AI Transparency Act, are on our AI content disclosure hub.
A law enforcement sweep the FTC announced on September 25, 2024, with five actions against operations it said used AI hype or sold AI technology in deceptive and unfair ways, including a service marketed as a robot lawyer.2
Not across the board. The FTC set aside its 2024 Rytr order on December 22, 2025, citing the federal AI Action Plan, while approving a final order over AI accuracy claims in August 2025 and settling with Air AI in March 2026.4
You update the approved version once. Arbitr checks what's already live against it and shows you every version still using the old wording.
