Key Takeaways
- The FTC proposes that any AI output steered toward a goal the user did not expect – even to comply with anti‑discrimination or state laws – is likely deceptive under Section 5 of the FTC Act.
- State laws that force such adjustments do not provide a safe‑harbor; the FTC argues they may be pre‑empted because they conflict with the federal prohibition on deceiving consumers.
- A company’s motive (profit, public pressure, employee politics, or legal compliance) is irrelevant; if deception occurs, liability follows.
- To avoid deception, firms must make clear, prominent disclosures whenever they steer outputs away from what users reasonably expect; burying notices in terms of service or fine print is insufficient.
- Ordinary AI errors (hallucinations) and blocking illegal content are not per se deceptive, but overstating a system’s reliability can still violate Section 5.
- Comments on the proposal are due July 31, 2026 (Docket No. FTC‑2026‑0859).
Background and Legislative Context
On July 7, 2026, the Federal Trade Commission released a proposed policy statement explaining how it intends to regulate output from artificial‑intelligence systems. The statement was issued in response to Executive Order 14365, “Ensuring a National Policy Framework for Artificial Intelligence” (Dec. 11, 2025), which directed the FTC to clarify how state laws that compel changes to AI outputs might conflict with federal law. The FTC also ties its analysis to a separate order concerning fair‑lending enforcement, noting that the administration rejects disparate‑impact liability under fair‑lending statutes. As the agency explains, “The FTC’s position is simple. If an AI company steers its system to produce answers that serve some goal other than what the user expected, the company is likely deceiving its customers in violation of federal law – specifically, Section 5 of the FTC Act, which prohibits unfair or deceptive acts or practices, or UDAPs.”
The FTC’s Theory of Deception
The FTC bases its approach on its 1983 Policy Statement on Deception, which holds that a practice is deceptive when it is likely to mislead a reasonable consumer and the misleading claim matters to the consumer’s decision. The agency emphasizes that AI marketing routinely promises that systems will give “the most accurate answer possible,” creating an expectation that outputs will be truthful and aligned with the user’s own goals. Citing internal data, the FTC notes that “consumers accept AI answers without fact‑checking more than 90 % of the time.” Consequently, steering a model toward hidden objectives—such as fitting an ideological agenda, advancing equity goals, or avoiding politically sensitive topics—constitutes a material misrepresentation because consumers rely on those promises. The proposal acknowledges that legitimate model goals (brevity, clarity, relevance, accuracy) and user‑requested creative inaccuracies are permissible, but any undisclosed deviation from what users expect is problematic.
Conflict with State Law
The most consequential portion of the proposal addresses state AI statutes, singling out Colorado’s revised Artificial Intelligence Act (S.B. 26‑189). The FTC asserts that Section 5 offers no state‑law safe harbor: “State law that requires an AI firm to deceive its consumers obviously conflicts with [FTC Act] Section 5’s express purpose of protecting consumers from such conduct.” According to the statement, if a company alters its outputs to satisfy a state mandate—whether to avoid discrimination, comply with equity requirements, or meet other policy goals—it may still be deemed deceptive unless it provides a clear, prominent disclosure. The FTC further argues that such state laws are likely pre‑empted because they frustrate the federal objective of preventing consumer deception. This stance sets up a direct clash between federal and state authority, warning firms that compliance with Colorado‑style statutes could expose them to FTC enforcement unless they adequately inform users.
What This Means for Companies
The policy statement applies not only to developers of foundational AI models but also to any organization that deploys consumer‑facing AI, including those that rely on third‑party systems. Under third‑party risk‑management principles, the compliance burden of Section 5 can extend to service providers that integrate AI into their products. Companies should therefore inventory the goals their models are optimized to achieve, compare those goals with the accuracy, objectivity, and capability claims made in marketing and product documentation, and determine whether any gap exists. If a discrepancy is found, the FTC advises that “clear, prominent disclosure” is required; disclosures hidden in terms of service or shown only once in fine print will not satisfy the standard.
For firms subject to state AI laws—particularly Colorado’s revised Act—the guidance creates tension: state mandates may push firms toward output adjustments that the FTC views as deceptive unless prominently disclosed. Until courts resolve the pre‑emption question, businesses are encouraged to document their compliance choices and consider whether transparent disclosure can satisfy both regimes.
Practical Steps and Comment Process
The proposal outlines concrete actions for stakeholders: review marketing claims about accuracy and objectivity, ensure those claims can be substantiated, and implement conspicuous notices whenever output steering occurs. The FTC stresses that ordinary technical mistakes—often termed “hallucinations”—do not automatically trigger Section 5 liability, but exaggerating a system’s reliability could still be deceptive. Blocking illegal content or thwarting cyberattacks remains permissible.
Interested parties have until July 31, 2026 to submit comments via regulations.gov, referencing Docket No. FTC‑2026‑0859 and the matter number P264200. The agency invites input on the definition of adequate disclosure, what constitutes an “expected objective,” and the pre‑emption analysis, all of which could shape the final version of the policy statement.
Conclusion
The FTC’s proposed policy statement signals a stringent stance: any undisclosed steering of AI outputs toward goals users did not anticipate—whether driven by profit, ideology, employee sentiment, or state‑law compliance—is likely to be deemed deceptive under Section 5. While the proposal leaves room for legitimate model objectives and user‑requested creative deviations, it places a high burden on firms to disclose any divergent practices clearly and prominently. The debate over federal pre‑emption of state AI laws is poised to intensify, making the upcoming comment period a critical opportunity for industry to influence the final rule. As the FTC succinctly puts it, “If an AI company steers its system to produce answers that serve some goal other than what the user expected, the company is likely deceiving its customers.” Companies that heed this warning and adjust their disclosures accordingly will be best positioned to avoid enforcement risk in the evolving AI regulatory landscape.
https://www.spencerfane.com/insight/ftc-proposes-new-policy-on-ai-accuracy-hiding-how-an-ai-system-is-steered-may-violate-federal-law/