FTC Warns Hidden Ideological Steering of AI Chatbots May Violate Federal Law

The US Federal Trade Commission (FTC) published a proposed policy statement on July 1, 2026, stating that AI companies that intentionally skew their chatbots' responses toward undisclosed ideological objectives may be violating Section 5 of the FTC Act, which prohibits unfair or deceptive business practices. The agency has opened a public comment period running through July 31, 2026.

The FTC's position rests on the premise that users of AI chatbots have a reasonable expectation that these systems aim to provide truthful and accurate answers. If a company trains its model to avoid responses perceived as discriminatory toward certain groups, or otherwise deliberately skews outputs away from the most accurate answer without clearly informing the user, this could constitute an unfair or deceptive practice under federal law.

Customers have a reasonable expectation that AI systems strive to provide truthful and accurate responses - US Federal Trade Commission, proposed policy statement

The Executive Order Behind the Policy

The FTC's proposal did not emerge in a vacuum. Its legal basis is Presidential Executive Order 14365 of December 11, 2025, which directed the agency to clarify how Section 5 of the FTC Act applies to artificial intelligence models, particularly in the context of state laws that mandate modifications to AI-generated responses. In the document, the FTC explicitly cites Colorado's artificial intelligence law as an example of regulation that could push companies to suppress the accuracy of their systems in order to avoid liability for disparate impact on different groups of users.

This illustrates the broader political dispute in the United States over whether state laws designed to protect against discriminatory AI outputs might themselves be forcing companies to falsify model responses. The FTC's position is that regardless of motivation, a lack of transparency toward the user is a legal problem.

How Companies Can Avoid Liability

The proposal outlines a clear path for AI companies to avoid liability. They can avoid violating the law if they clearly, visibly and adequately inform users that their system prioritizes certain objectives over what the user would expect or request. The FTC specifies that a disclosure buried in terms of service or fine print is not sufficient. The further a model's behavior deviates from what a user would reasonably expect, the more visible the warning must be.

This distinction carries practical weight for chatbot developers. A model that openly states its moderation limits in the user interface is in a different legal position than one that silently filters or reframes responses without any mention of it during the conversation with the user.

The Global Context

The American approach, based on transparency and disclosure, contrasts with the EU's regulatory model, which relies on hard financial penalties. Under the EU AI Act, the maximum penalty for non-compliance reaches 35 million euros or 7 percent of a company's global annual turnover, and the main wave of enforcement for high-risk systems takes effect on August 2, 2026. Since 2022, regulators and courts in the European Union have already imposed fines exceeding 13 billion zloty on seven technology companies across ten separate AI-related cases, mostly concerning the use of data to train models in violation of EU rules.

For Polish users and businesses using chatbots such as ChatGPT, Gemini or Claude, the FTC's proposal has only indirect relevance, since it concerns the US market, but it could influence the global practices of these tools' providers. Companies operating in both the US and the EU will need to reconcile the American transparency and disclosure requirements with EU rules on high-risk systems and potential financial penalties, which could translate into more visible notices about model limitations even in interfaces available in Poland.

What Comes Next

Until July 31, 2026, the FTC is collecting comments from companies and consumers about their experiences with AI systems they believe have been shaped to serve undisclosed ideological goals. Once the consultation ends, the agency could adopt the statement in final form, opening the door to enforcement actions against companies that fail to adjust their disclosure practices. Critics point out that the draft statement leaves unclear the line between "hidden ideological steering" and the standard content moderation practiced by every major language model provider.

Sources: US Federal Trade Commission (ftc.gov), Consumer Financial Services Law Monitor (consumerfinancialserviceslawmonitor.com), Rzeczpospolita (rp.pl)