FTC warns AI firms: hidden output steering may violate law
FTC targets undisclosed AI output steering
The Federal Trade Commission released a proposed policy statement on July 1, 2026, concerning what it calls "suppression of accuracy" in artificial intelligence systems. The proposal establishes that AI companies steering model outputs toward undisclosed objectives—without telling users—may face enforcement action under Section 5 of the FTC Act, which prohibits unfair or deceptive practices.
The regulatory position is direct: "steering a model's outputs toward any other undisclosed objective is therefore misleading under Section 5 of the FTC Act," according to the FTC's language in the proposal. The agency framed the issue as consumer protection, arguing that when an AI system steers away from what users reasonably expect without disclosure, it crosses into deceptive territory.
What the policy covers
The scope extends beyond accuracy metrics alone. The FTC's framework addresses any hidden steering—whether toward ideological objectives, safety constraints, or commercial interests—if that steering diverges from user expectations and goes undisclosed. "AI companies that steer model outputs toward undisclosed ideological objectives may violate Section 5 of the FTC Act's ban on deceptive practices," the agency warned in its proposal materials.
The proposal carries Matter No. P264200 and applies the FTC's existing deception and unfairness standards to a novel domain: the alignment and steering mechanisms that shape what AI systems output. The policy does not criminalize all output steering—only that which is undisclosed and reasonably unexpected by consumers or enterprise users.
Comment period and next steps
The FTC opened a public comment window and closed it on July 31, 2026, giving AI vendors, developers, and affected parties 30 days to weigh in. The agency has not yet published enforcement guidance or final rules; the July 1 release was a proposal intended to signal regulatory direction and invite input before the FTC finalizes its stance.
For agent builders and AI product teams, the immediate implication is clear: any output steering mechanism must either be transparent to users or grounded in disclosed safety or service policies. Hidden optimization away from accuracy or truthfulness invites regulatory scrutiny.