The Federal Trade Commission has opened a new front in Washington's campaign to rein in state AI regulation. In a proposed policy statement published in the Federal Register on July 7, the agency argues that secretly steering an AI model's outputs away from accurate answers — whether a company does it on its own or to satisfy a state law — can amount to deception under Section 5 of the FTC Act. Public comments are due July 31, and the stakes reach well beyond the docket.
What the FTC Is Proposing
The document, formally titled the Policy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems, was approved for publication by a 2-0 Commission vote and issued on July 1. It responds to Executive Order 14365, signed in December 2025, which directed the FTC to clarify how its deception authority applies to AI models and to address state laws the administration says force companies to alter truthful model outputs.
The core theory is straightforward. When consumers use an AI assistant, the agency reasons, they expect answers shaped by the model's best assessment of the facts. If a developer configures a system to pursue undisclosed ideological objectives — nudging outputs toward preferred conclusions without telling users — that gap between expectation and reality can be a deceptive practice under federal consumer protection law.
The statement draws some notable boundaries:
- Ordinary hallucinations — mistakes that stem from technical limitations rather than intentional design choices — do not, by themselves, raise Section 5 concerns.
- Companies can avoid liability through clear and conspicuous disclosure that a system prioritizes particular objectives, though the FTC cautions that a line buried in terms of service will not cut it.
- Compliance with a state statute is no defense. The agency asserts that conflicting state laws are impliedly preempted by the federal scheme.
The State Law Collision
That last point is where the proposal becomes politically explosive. The statement singles out Colorado's Artificial Intelligence Act — the first comprehensive state law targeting algorithmic discrimination — as an example of legislation that could pressure developers to suppress accurate outputs in the name of bias mitigation. Legal commentators note the logic could sweep further, potentially touching employment discrimination provisions in Illinois and California that now expressly cover AI systems.
The FTC move does not stand alone. Executive Order 14365 also created a Justice Department task force charged with challenging state AI laws in court beginning this year, and directed the Commerce Department to weigh state AI rules when distributing billions in broadband funding. Meanwhile, Congress is circling the same question: a bipartisan discussion draft released in June, the Great American Artificial Intelligence Act, would preempt certain state AI development rules for three years.
States, for their part, are not slowing down. Trackers count more than one hundred AI-related state laws enacted in 2026 by midyear, with activity clustering around child safety, data centers, and consumer protection — categories the federal push has largely carved out.
What Happens Next
A policy statement is not a rule. It creates no new legal obligations and binds no court; it signals how the Commission intends to wield its existing deception authority. But signals matter. AI companies now face a genuine dilemma: a state regulator may demand bias mitigation measures that a federal agency characterizes as unlawful output manipulation. Until courts resolve the preemption question, most counsel are advising clients to keep complying with state law while documenting design decisions and disclosures carefully.
The comment window closes on July 31, and submissions from AI developers, civil rights groups, and state attorneys general are expected to preview the litigation arguments to come.
Why It Matters
The deepest question in U.S. AI governance right now is not what the rules should say but who gets to write them. With Congress yet to pass a comprehensive AI statute, states filled the vacuum — and the executive branch is now testing every available lever to pull authority back to Washington. The FTC's accuracy statement is the most creative lever yet, recasting a consumer protection tool as a shield against state mandates.
For businesses, the practical takeaway is disclosure. Whatever a model is optimized for — safety, brand voice, legal compliance, ideology — the emerging federal expectation is that users be told plainly. Companies that build transparent disclosure into their AI products today will be better positioned no matter which sovereign wins the tug-of-war.
