FTC Weighs a Policy Targeting AI Systems That Suppress Accuracy

FTC Weighs a Policy Targeting AI Systems That Suppress Accuracy

The Federal Trade Commission has put artificial intelligence companies on notice that tuning a chatbot to quietly serve a hidden agenda could run afoul of federal consumer-protection law. In a proposed policy statement released July 1 and published in the Federal Register on July 7, the agency laid out a theory under which marketing an AI system as accurate and objective while secretly steering its outputs may amount to a deceptive act or practice under Section 5 of the FTC Act. The public has until July 31, 2026, to comment.

The document, titled "Proposed Policy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems," was authorized on a 2-0 Commission vote. It marks one of the more assertive federal moves on AI in a year otherwise defined by a deregulatory posture in Washington and a widening patchwork of state laws.

What the statement actually does

The proposal does not create new rules or ban any specific technique. Instead, it applies the FTC Act's long-standing prohibition on "unfair or deceptive" conduct to the AI marketplace. The core argument: AI developers have made explicit and implicit representations that their systems deliver accurate, objective, and useful outputs. If a company then intentionally designs a system to pursue undisclosed objectives that diverge from what users reasonably expect, the FTC says, that gap between promise and product can constitute deception.

The statement lists examples that could raise concerns, including modifying outputs to advance ideological goals, changing outputs in response to political or public pressure, or altering outputs to limit liability under certain state laws. It draws a sharp line between this kind of deliberate "steering" and AI "hallucinations," noting that inaccuracies stemming from technological limitations generally do not raise the same concerns. Companies can mitigate exposure through disclosures, the FTC adds, but only if they are clear and conspicuous, not buried in terms of service or fine print.

One clarification matters for reading the document correctly. As attorney Kevin Moriarty argued in a July 23 analysis for Tech Policy Press, the legal theory targets statements a company makes about its models, not the statements produced by the models themselves. Under this framing, a developer remains free to build an AI with a particular slant, provided it does not represent, explicitly or implicitly, that the system is free of one. Covington & Burling lawyers Laura Kim and Carter McCants, writing on the firm's Inside Privacy blog, similarly noted the proposal breaks from earlier FTC efforts that focused on false claims about AI capabilities, shifting instead to whether systems pursue undisclosed objectives.

Why an executive order is behind it

The statement did not emerge in a vacuum. It was issued pursuant to Executive Order 14365, "Ensuring a National Policy Framework for Artificial Intelligence," which President Trump signed on December 11, 2025. That order directed the FTC to explain when state laws requiring alterations to the "truthful outputs of AI models" are preempted by the FTC Act's deception prohibition. It singled out Colorado's Artificial Intelligence Act as a law that the administration argued could push developers to embed bias into their systems, though Colorado subsequently repealed and revised that statute.

Chairman Andrew N. Ferguson framed the initiative squarely in the administration's terms. "The FTC wants to hear from businesses and consumers about their experiences and concerns regarding the subversion of AI systems for ideological ends," he said. "This 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."

The tension beneath the policy

Here is the paradox worth flagging. The order that produced this statement is part of a broad deregulatory push meant to clear state-level obstacles for AI firms. Yet the statement itself stakes out an expansive view of the FTC's authority to police AI companies under its deception power. Analysts have noted the two impulses do not sit comfortably together.

Moriarty concluded that the statement "stakes out an aggressive approach to policing implied representations by AI companies" while failing to make a convincing case for preempting any state law. Because the FTC Act contains no express preemption, the agency leaned on a contested "obstacle" theory of conflict preemption, an argument several conservative jurists have criticized. The upshot, in his reading, is a document that may do little to lighten the regulatory load on industry and instead lays out a "road map for aggressive enforcement" that this and future Commissions could follow.

That the vote was 2-0, with no dissenting commissioner of the opposing party on the current Commission, also means the proposal drew no formal internal counterargument, leaving the public comment process as the main channel for pushback.

What to watch

The immediate marker is the July 31 comment deadline; filings can be submitted through the docket at Regulations.gov. Beyond that, watch whether the FTC moves from policy statement to actual enforcement, and whether it brings a case testing the "suppression of accuracy" theory against a major developer. The preemption claims are likely to invite litigation if the agency ever tries to use them to knock down a state AI law. And because the deception logic here is broad, the same reasoning could extend well past ideological bias, to implied promises about data use, security, and unbiased product recommendations. For an agency that spent 2026 mostly stepping back, this is a notable signal that the FTC still intends to be an AI enforcer.