Source: Arnold & Porter (Major Questions blog)

The Federal Trade Commission issued a proposed policy statement relying on Section 5 of the FTC Act to address the so-called “suppression of accuracy” in AI outputs. The FTC would consider unfair or deceptive acts or practices those AI companies that allegedly steer system outputs toward undisclosed objectives, including ideological agendas, rather than the user-requested or reasonably expected objectives. Public comments are invited on the proposed statement, with a deadline of July 31, 2026. The proposal follows Executive Order 14365, which instructed the FTC to clarify the preemptive effect of Section 5 over state laws requiring alterations of truthful AI outputs. The statement discusses state AI laws generally and references Colorado’s revised Artificial Intelligence Act as an example of state legislation that may lead to output distortions.

If adopted, the proposal may expose not only AI model developers but also downstream entities that create systems based on those models and customers who fine-tune or deploy them to enforcement proceedings, as any perceived bias or steering inconsistent with user expectations could trigger FTC scrutiny. The FTC distinguishes undisclosed, intentional “distortion” of outputs from ordinary technological “hallucinations,” which it does not associate directly with Section 5 violations, although misrepresentations about hallucinations may be actionable. The policy statement emphasizes that adequate and prominent disclosures about model objectives may prevent liability. It also states that compliance with state law does not excuse undisclosed distortion and that state laws may be impliedly preempted where they conflict with Section 5’s prohibitions. The FTC specifically notes concerns that state laws, such as Colorado’s statute, may pressure AI companies to modify outputs in ways that depart from user expectations and thus potentially constitute deceptive practices under Section 5.