FTC Proposed Policy Statement on AI Output Accuracy: Issues in Performance Claims and Consumer Protection

On July 1, 2026, the U.S. Federal Trade Commission (FTC) published the Proposed Policy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems and opened a public comment period. The proposed statement sets out the view that an AI business steering the output of an AI system toward a particular ideological goal without disclosing this to consumers may constitute a deceptive act under Section 5 of the FTC Act. In particular, the main scenario envisaged is adjusting output content in response to state laws such as Colorado's AI law. What can be confirmed as of the time of writing (July 12, 2026) is a proposal subject to public comment, not a finalized regulation. This article organizes the legal nature of what was published and the practical issues that companies involved in AI services should check.

What Was Published

In a press release of the same date, the FTC announced that it would publish the proposed policy statement in the Federal Register and seek comments. The deadline for comments is July 31, 2026, and comments are to be submitted through regulations.gov. As of the time of writing, the deadline has not passed and the comment period is ongoing. The press release states that the Commission vote approving publication in the Federal Register was 2–0, but I think this was a vote approving the procedure of publication in the Federal Register and does not indicate approval or disapproval of the content of the proposed statement itself. As of the time of writing, no written concurring or dissenting statements by individual commissioners on the content of the proposed statement itself could be found.

The proposed statement is set against the background of Executive Order 14365 of December 11, 2025 ("Ensuring a National Policy Framework for Artificial Intelligence"). That Executive Order directed the FTC to issue a policy statement on enforcement clarifying how state laws may conflict with Section 5 of the FTC Act where they require alterations to the accurate outputs of AI models.

Legal Nature of the Policy Statement and Section 5 of the FTC Act

A policy statement is understood to be a document in which the FTC publicly sets out its future approach to the interpretation and application of an existing statutory enforcement power, Section 5 of the FTC Act, and does not itself create new regulations or laws. The proposed statement applies to AI systems the framework for determining deception set out in the Policy Statement on Deception published by the FTC in 1983. That framework consists of three elements: there is a representation, omission or practice that is likely to mislead consumers; it is assessed from the perspective of a consumer acting reasonably; and the representation or the like is material, affecting consumers' decisions. The proposed statement itself also sets out the analysis that implied preemption under Section 5 of the FTC Act may apply in relation to state laws, including Colorado's amended AI Act.

The Approach Set Out in the Proposed Statement

The central claim of the proposed statement is that, because AI businesses have expressly or impliedly represented their AI systems as products that aim for the "best output," achieving the user's desired goal as faithfully and accurately as possible, consumers have a reasonable expectation that the output is not distorted by an undisclosed, particular ideological goal. The proposed statement refers to Colorado's revision, by an amending law enacted on May 14, 2026, of the Artificial Intelligence Act it enacted in 2024, and states that a business prioritizing other goals, such as avoiding discriminatory outcomes or "fairness," while sacrificing the accuracy of output without disclosing that fact may constitute a deceptive act under Section 5 of the FTC Act, even where it does so for the purpose of complying with state law.

On the other hand, the proposed statement expressly states that so-called "hallucinations," that is, erroneous output itself resulting from technical or resource constraints, do not in themselves give rise to issues under Section 5 of the FTC Act; the approach applies only to intentional and undisclosed manipulation of output and is distinguished from the technical limits of performance. In addition, at this stage the proposed statement confines its analysis to the issue of deception and reserves judgment on whether such conduct is unfair. As background, the proposed statement also refers to past enforcement actions based on claims about accuracy and performance, such as accuracy claims for AI content detection tools, claims that AI chatbots can replace professional services, and performance claims for facial recognition and security detection systems, and the general principle under Section 5 of the FTC Act that performance and accuracy claims require substantiation itself is maintained. As for disclaimers, the proposed statement states that a business may avoid deception by disclosing that its design prioritizes a particular goal, but that such disclosure is not sufficient if buried in the fine print of the terms of use; it must be made clearly and conspicuously, and to a degree sufficient to overcome consumers' expectations.

Points of Contact for Japanese Companies

For Japanese companies providing AI services to the U.S. market, the starting point is to identify whether their English websites, product descriptions and terms of use contain any express or implied representations about the objectivity, neutrality or accuracy of their services. Where a company's service adopts a design that intentionally adjusts part of its output based on regulatory compliance in a particular jurisdiction or on internal policy, it is necessary to check to what extent and where that policy is disclosed to U.S. users, and to examine whether disclaimers are placed only in inconspicuous parts of the terms of use. Perspectives for reviewing terms of use are organized in The Legal Issues to Look at First in the Terms of Use of Generative AI Services.

Separately, for numerical claims about accuracy and performance used in advertising and the like, it is necessary to check whether the company holds substantiating materials. This idea partly overlaps with the requirement to hold reasonable substantiating materials under the regulation of misleading representations in the Premiums and Representations Act for representations directed at Japan, but the two are different regimes of different countries; confirming the lawfulness of representations directed at the United States does not mean the company has fulfilled its obligations under Japan's Premiums and Representations Act, and vice versa. After organizing which representations the company makes to consumers in which countries, it is necessary to check both regimes in parallel without confusing them. For items to check when using overseas AI vendors, see Checkpoints for AI Vendor Due Diligence and Contract Review; for legal checks in general when providing AI services, see The First Legal Checklist a Company Starting an AI Service Should Create; and for an overall picture of Japan's AI-related guidelines, see Checkpoints for Responding to the AI Promotion Act and the AI Guidelines for Business.

Developments to Watch

As of the date of writing, the following points remain unsettled. The comment period continues until July 31, 2026, and it is not clear what revisions the FTC will make before publishing a final version. How the relationship between Colorado's amended AI Act, on which the proposed statement is premised, and the concept of preemption under Section 5 of the FTC Act will be treated in future litigation is also a subject to monitor. If a final policy statement is published, the starting point will be to compare it with the content of the proposal organized in this article and check how it responds to the issues raised in the comments, the specific standards for disclosure, and whether the scope of application has been clarified.