The FTC’s Proposed Policy Statement Concerning The Suppression Of Accuracy In Artificial Intelligence Systems

On July 1, 2026, the Federal Trade Commission (FTC) issued a proposed policy statement addressing how Section 5 of the FTC Act applies to AI systems. The statement was issued pursuant to Executive Order 14365, signed by President Trump on December 11, 2025, which directed the FTC to clarify how its deception framework applies to AI models and to address potential conflicts between federal law and state laws requiring alterations to the accurate outputs of AI models. The statement reflects the current Administration’s broader goal of establishing a national AI regulatory framework that preempts state-by-state regulation.¹ The public comment period runs through July 31, 2026.

This is a proposed policy statement, not a binding regulation, which alone will not create new legal obligations. However, it provides a glimpse at how the Commission might apply its existing deception framework to AI systems, and companies developing or deploying AI products whose outputs reach consumers should take note when evaluating the scope and adequacy of their AI disclosures.

Background

The FTC takes the position that AI companies have spent years representing, whether explicitly or implicitly, that their systems aim to produce the best, most accurate output possible within technological and resource constraints. The FTC cites data suggesting consumers accept AI outputs without independent fact-checking more than 90% of the time. Given those representations and consumer reliance, the FTC notes that AI companies could, for instance, be tempted to exploit consumer trust by covertly training a model to generate ideologically driven distortions when answering factual questions, such as altering historical facts to align with the developer’s view of correcting perceived historical injustices. Similarly, a company could face pressure from state laws or government mandates to modify the outputs produced by its technology. Here, the FTC observed that the original version of Colorado’s Artificial Intelligence Act broadly obligated AI companies to avoid output that might lead to disparate impacts in various contexts. The FTC notes that such distortions would violate Section 5 of the FTC Act.

Although Colorado was specifically referenced in the policy statement, it is likely not the only state whose laws could potentially conflict with obligations under the FTC Act. The Illinois Human Rights, for example, imposes liability for discriminatory AI outputs in the employment context,² California’s Fair Employment and Housing Act was recently amended to apply to the use of artificial intelligence,³ and Executive Order 14365 itself identified Colorado as an example of a state whose law could pressure AI companies to alter their systems’ outputs to avoid disparate impact on protected groups. On February 25, 2026, the Connecticut Attorney General issued a memorandum noting that “antidiscrimination laws apply to algorithmic discrimination in the same way they have long applied to other discriminatory conduct irrespective of whether discriminatory conduct is facilitated by automated decision-making tools or driven by purely human practices.”⁴

The policy statement applies the FTC’s three-part test for deception: a representation, omission, or practice that is likely to mislead a consumer acting reasonably under the circumstances, and that is material to the consumer’s decision.

Intentional steering vs. hallucinations

The statement draws an important line between deliberate design choices and technical errors. AI “hallucinations”—incorrect outputs stemming from technological limitations rather than deliberate design choices—do not, by themselves, raise Section 5 concerns. Additionally, the statement does not prohibit companies from blocking content clearly unprotected by the First Amendment (such as child pornography) or imposing limits to prevent cybersecurity attacks. 

Impact of state laws

Part of the statement is directed at examining state AI laws, with Colorado’s Artificial Intelligence Act used as an example, although other states, including Illinois and California, have adopted or proposed similar legislation.

The FTC describes Colorado’s law as one that could pressure companies into altering outputs to avoid disparate impact liability. The revised law explicitly states that both a developer and a deployer can be held liable for discriminatory outcomes caused by their customers’ use of their AI products.⁵ The FTC views it as predictable that an AI company might suppress accuracy and interpose other objectives, what the FTC describes as “so-called ‘equity’” to avoid liability under this type of law, but fail to disclose these altered objectives to consumers. 

Notably, the statement takes the position that compliance with a state law is not a defense to a federal deception claim. The FTC argues that such state laws are “impliedly preempted to the extent [they] conflict[] with a federal regulatory scheme.” A state law that requires an AI firm to deceive its consumers, in the FTC’s view, conflicts with Section 5’s express purpose of protecting consumers from deceptive conduct. Here, the FTC emphasizes that Congress did not provide state-law safe harbors under Section 5. 

Practical implications for companies

Companies subject to state AI laws should evaluate whether steps taken to comply with those laws could be characterized as intentional output suppression that conflicts with consumers’ reasonable expectations and, in turn, with Section 5. This creates direct tension for companies operating in multiple jurisdictions, as compliance with one regime may create exposure under the other.

Although the policy statement refers broadly to “AI companies” without distinguishing between developers and deployers, the Commission’s recent enforcement history suggests that entities making consumer-facing representations about AI capabilities may face Section 5 liability regardless of whether they built the underlying model. For example, the Commission's 2025 final order against DoNotPay targeted deceptive claims about the capabilities of an AI-powered service, focusing on the company's representations to consumers rather than on whether it developed the AI technology itself.⁶

Disclosure as a safe harbor

The statement acknowledges that an AI company can avoid Section 5 liability by making clear, conspicuous, and adequate disclosures that its system is designed to prioritize certain objectives over what users request or would otherwise expect.

• The FTC outlines the following guidelines for disclosures:

• A disclaimer buried in terms of service would not suffice.

• The degree of prominence required scales with how far the disclosure departs from what a user would otherwise expect. 

• Representations would need to make clear that the AI company is prioritizing objectives different from those consumers requested or would otherwise expect. 

In practice, calibrating these disclosures presents a significant challenge. Companies will need to make product- and use case-specific judgments without the benefit of detailed regulatory guidance. Winston’s AI Group regularly advises clients on designing disclosure practices that are defensible under evolving federal and state standards.

Footnotes

1. In a July 1 press release, Commission Chairman Ferguson noted that “[t]his crucial input will help the Commission formulate a final policy that advances President Donald Trump’s goal of expanding America’s global dominance in artificial intelligence.” FTC Seeks Public Comment on Policy Statement Addressing AI Accuracy | Federal Trade Commission

2. 775 ILCS 5/2-102.

3. Cal. Code Regs. tit. 2, §§ 11008, 11008.1 (“Automated-Decision Systems).

4. Connecticut Office of the Attorney General, AI Advisory (2026), https://portal.ct.gov/-/media/ag/press_releases/2026/office-of-the-attorney-general---ai-advisory.pdf.

5. Colo. Sen. Bill 24-205 (enacted May 17, 2024), repealed and reenacted, as amended, by Colo. Sen. Bill 26-189 (enacted May 14, 2026).

6. Fed. Trade Comm’n, FTC Finalizes Order with DoNotPay That Prohibits Deceptive “AI Lawyer” Claims, Imposes Monetary Relief, and Requires Notice to Past Subscribers (Feb. 11, 2025), https://www.ftc.gov/news-events/news/press-releases/2025/02/ftc-finalizes-order-donotpay-prohibits-deceptive-ai-lawyer-claims-imposes-monetary-relief-requires.