FTC Issues Proposed Policy Statement On AI Accuracy And Output Steering

On July 1, 2026, the Federal Trade Commission (FTC) issued a proposed policy statement focusing on AI models that steer outcomes toward undisclosed ideological objectives. According to the FTC, this practice may compromise accuracy and be contrary to reasonable consumer expectations about how AI models work. The proposed policy statement follows Executive Order 14365, which specifically directs the agency to clarify how § 5 of the FTC Act applies to both AI models and emerging state laws that regulate those models.

Section 5 Analysis

According to the proposed policy statement, consumers generally expect AI systems to strive for accurate responses and faithfully carry out consumer requests, subject to ordinary technological limitations. These expectations are reasonable, the FTC contends, due to the inherent nature of AI systems and explicit or implied representations that companies have espoused for years about those systems. Unbeknownst to consumers, however, AI developers may engage in “steering” the output, the FTC says, by deliberately configuring AI systems to advance objectives other than accuracy (e.g., achieving ideological goals, avoiding controversial topics, or attempting to limit legal exposure). 

To illustrate this point, the proposed policy statement specifically points to Colorado’s recently enacted Artificial Intelligence Act, which provides, among other requirements, that companies can be held liable for discriminatory AI model outputs resulting from customers’ use of those products. The FTC asserts that even a company’s efforts to comply with this statute (and similar laws in other states) could compromise accuracy and be contrary to consumers’ reasonable expectations about how AI models work. In the FTC’s view, failure to sufficiently disclose this to consumers may violate § 5 because deception under the FTC Act includes “implied misrepresentations, half-truths, and instances where only a ‘significant minority’ of consumers are misled.”

The proposed policy statement appears to set a high bar as to what constitutes an adequate disclosure on outcome steering, noting that the disclosure cannot be buried in the terms of service, and a one-time disclosure “subsequently hidden away in fine print” is unlikely to suffice. Instead, the disclosure must “clearly and conspicuously dispel the notion that the system is designed to give the best answer possible.” Additionally, the FTC clarifies that the more the disclosure cuts against the reasonable expectations that users would take away from other contexts, the more persistent and prominent the disclosure would need to be.

Tension with State Laws

In the proposed policy statement, the FTC considers how emerging state laws on AI, including the Colorado Artificial Intelligence Act, may require AI developers to alter, constrain, or otherwise influence AI outputs to achieve particular public policy objectives. The FTC takes the position that compliance with a state requirement does not automatically eliminate potential federal deception concerns if consumers are not adequately informed about how those requirements affect outputs.

The proposed policy statement also suggests that state laws compelling modifications to AI outputs could, in some instances, conflict with federal consumer protection principles. Although the proposed policy statement does not provide a formal preemption rule, its discussion of emerging states AI laws seems to reflect the current administration’s interest in promoting a more uniform national framework for AI regulation.

Takeaways for Businesses

The proposed policy statement provides a tentative roadmap to how the FTC may analyze questions involving AI governance, transparency, and content controls in the future. Stakeholders should consider reviewing the content of their marketing communications and other public-facing statements — not only for explicit claims about AI products, but also for implied representations or potentially relevant information that has been omitted. Companies should consider assessing governance processes for AI model training and content moderation, given the FTC’s apparent scrutiny of intentional design choices. They should also consider how to address the potential disconnect between state AI and privacy laws (and international laws, such as the EU AI Act, where relevant) and the FTC’s position.