The Federal Trade Commission (FTC) has repeatedly emphasized that there is no “AI exemption” from consumer protection laws. Recent enforcement actions have focused on AI washing, exaggerated performance claims and misleading representations about AI capabilities.

The FTC’s July 1, 2026 proposed policy statement goes a step further. Companies may violate Section 5 of the FTC Act if they intentionally steer AI outputs toward undisclosed objectives that differ from users’ reasonable expectations regarding accuracy, objectivity or truthfulness. If your business is simply using AI to help create content or streamline workflows, the FTC’s proposal is aimed elsewhere. This proposal’s primary focus is on the companies that build, train and control AI systems – not the businesses that use them.

The proposal follows Executive Order 14365, which directed the FTC to clarify how existing consumer protection laws apply when AI systems generate outputs that may conflict with user expectations. The proposal does not create new legal obligations, but it outlines conduct that the FTC believes could violate existing Section 5 deception principles. Specifically, the FTC suggests that:

  • AI companies should not secretly steer outputs toward objectives that differ from those requested by users or reasonably expected by consumers.
  • If an AI system prioritizes other objectives – such as ideological, political, policy-based or commercial goals – those objectives should be clearly disclosed.
  • Disclosures must be prominent and effective; disclosures buried in terms of service or fine print may not be sufficient to alter consumer expectations.
  • AI companies remain responsible for compliance even when output modifications are implemented to satisfy state law requirements.

What Types of Communications Are Covered?

The agency’s primary concern is that AI providers may influence outputs behind the scenes without adequately informing users.

Like in any deception case, the FTC’s analysis focuses on the overall message conveyed to consumers. The proposal applies to both express and implied claims about an AI system’s accuracy, truthfulness, objectivity, neutrality, reliability and intended purpose.

One of the clearest themes running through the proposal is transparency. If consumers are told that an AI system is designed to deliver accurate, objective answers but there is output steering or disclosures hidden in legal fine print, the AI company may face liability.

At its core, the FTC’s proposal targets situations where:

  1. Consumers use an AI system,
  2. The AI provider represents that the system is designed to provide accurate or objective responses,
  3. The provider intentionally steers outputs toward undisclosed objectives, and
  4. Users are not adequately informed of those objectives.

Notably, the proposal is directed primarily at companies that develop, train, fine-tune, deploy or control AI systems. Its examples focus on AI developers, model providers, chatbots and AI-powered services – not ordinary businesses using AI tools in their daily operations.

That does not mean businesses using AI are immune from FTC scrutiny. A retailer that uses AI to generate product claims or a beauty company that relies on AI-generated advertising could still face liability if the resulting claims are false or misleading. In those cases, however, the risk would arise from traditional advertising and consumer protection principles – not because the company is an AI provider allegedly steering system outputs.

Key Takeaways

The FTC’s message to AI companies is straightforward: transparency matters. If your system influences outputs in ways users would not reasonably expect, the FTC wants those practices clearly disclosed – not buried in the fine print.

AI companies should consider whether:

  • Output-ranking, filtering or steering practices are clearly documented and disclosed.
  • Marketing claims accurately describe how the system operates.
  • Statements regarding accuracy, neutrality, objectivity or reliability match the system’s actual design.
  • Users would understand when outputs are being shaped by objectives beyond simply providing the most accurate or the most responsive answer.

While the proposal is not yet final, it provides an early road map of the issues the FTC is likely to scrutinize as AI enforcement continues to evolve.