The FTC's Proposed AI Accuracy Policy Statement: What It Targets and Who It Covers

AI Breaking News is an AI-generated alert, curated and reviewed by the Kursol team. When major AI developments happen, we break down what it means for your business.

On July 1 the Federal Trade Commission proposed a policy statement saying AI companies that manipulate their systems' behavior contrary to consumer expectations for accuracy could be deceiving consumers under Section 5 of the FTC Act, on a 2-0 Commission vote. The statement published in the Federal Register on July 7, 2026, at 91 FR 41638 under File No. P264200, and comments must be received on or before Friday, July 31, 2026. It exists because Executive Order 14365, signed December 11, 2025, directed the Commission to clarify how state laws requiring alterations to the accurate outputs of AI models can conflict with the FTC Act.

What "Suppression of Accuracy" Actually Means

The core claim is narrower than the title suggests. The Commission says AI companies that "steer the outputs of their AI systems toward unexpected objectives, and away from the objectives set by or reasonably expected by users, are likely to deceive consumers in violation of section 5." The reasoning: vendors market their systems as producing the best output they can, consumers reasonably believe that, so a hidden competing objective makes the marketing claim false. As evidence of consumer reliance, the notice cites a Forbes write-up of an Anthropic finding that consumers accept AI outputs without further fact-checking over 90% of the time — the Commission's citation, not our measurement.

Two carve-outs matter more than the headline does. Hallucinations sit outside the statement: errors stemming from technological and resource limitations rather than a design decision do not, by themselves, raise Section 5 issues — though misrepresenting how often your system gets things wrong still can. And the Commission "at this time takes no position" on whether the same conduct is also unfair under the FTC Act, keeping a second enforcement theory in reserve.

The Named Target Is State AI Law

The FTC's press release names Colorado's Artificial Intelligence Act, saying it "appears to coerce companies into altering the output of their AI models to comply with and advance the state's ideological objectives," and argues such a law is "impliedly preempted to the extent it conflicts with a federal regulatory scheme." The Federal Register text is blunter on motive: a company's reasons are irrelevant to Section 5, so the prohibition applies "even when a company engages in a deceptive act or practice in order to comply with a State law." Chairman Andrew N. Ferguson said the FTC "wants to hear from businesses and consumers about their experiences and concerns regarding the subversion of AI systems for ideological ends."

Read the two documents side by side and the emphasis differs. The press release is about ideology and state law; the Federal Register text is about disclosure, and disclosure is the part with teeth for everyone else.

What This Means If You Deploy an AI Assistant

The statement is aimed at the companies that build and tune models, and that is the natural first reaction if you run a support chatbot on somebody else's model. But nothing in the document puts deployers outside its scope.

Your exposure sits in the second half of the document. It sets a disclosure standard: an adequate disclaimer "could not be buried in terms of service," would have to "clearly and conspicuously dispel the notion that the system is designed to give the best answer possible," and "it is doubtful a one-time disclosure subsequently hidden away in fine print would suffice." That standard reaches anyone putting an AI system in front of customers under their own brand.

The FTC already brings cases over AI accuracy claims. The notice cites its own recent actions on the performance and efficacy of AI products — consent orders involving DoNotPay's chatbot, IntelliVision's facial recognition software and Workado's AI content detection product, plus a complaint against Air AI Technologies over its conversational AI product. None involved ideology. Each involved what a company said its AI could do.

What to Do This Week

1. Audit what you promise your AI assistant does. Run every public claim against the FTC's three-part deception test: is it likely to mislead, would a consumer acting reasonably be misled, and is it material to their decision.

2. Write down every place you deliberately steer outputs. Content filters, refusal rules, brand-safety layers, prompts that push a preferred product. Then ask whether a customer reading your product page would expect them. Where the answer is no, the Commission's position is that the disclosure belongs somewhere prominent.

3. File a comment if this touches your product. The docket is at regulations.gov under FTC-2026-0859, with comments due July 31, 2026. If that window has closed by the time you read this, the docket stays public — the filed comments show how other companies read their own exposure, which is the better guide to what the Commission ends up adopting.

The Bottom Line

This is a proposed policy statement, not a rule, and the Commission says it will consider all timely and responsive comments received on or before July 31. What it marks is where FTC deception enforcement is heading: at the gap between what a company says its AI does and what it has quietly configured that AI to do.

For a mid-market business running a customer-facing AI assistant, the useful takeaway has nothing to do with the political fight driving the document. "Our AI gives you the best answer" is a marketing claim the FTC expects you to back up, and every rule you have layered on top of the model is a fact about that claim.

If you are not certain what your AI tools actually do, or what you have promised customers about them, take our free AI readiness assessment to see where you stand.

AI Breaking News is Kursol's rapid analysis of major artificial intelligence developments—focused on what actually matters for your business. Subscribe to our RSS feed to stay informed.

FAQ