On July 7, 2026, the Federal Trade Commission (“FTC”) issued a proposed policy statement addressing what it describes as the “suppression of accuracy” in artificial intelligence (“AI”) systems. The proposal would apply traditional deception principles under Section 5 of the FTC Act to AI systems whose outputs are allegedly steered away from users’ objectives or reasonable expectations regarding truthfulness and accuracy. The FTC requested public comment by July 31, 2026, approximately three weeks after publication of the proposal.
At a high level, the proposal focuses on circumstances in which an AI company’s statements or overall presentation create expectations about how the system will operate, but the system is designed or modified to prioritize undisclosed objectives that conflict with those expectations.
The proposed statement follows a December Executive Order directing the FTC to address the legal implications of state laws that require alteration of truthful AI model outputs. In the accompanying press release, Chairman Andrew N. Ferguson said the FTC sought input on “the subversion of AI systems for ideological ends.” Together, the executive order and proposed statement reflect the Trump Administration’s broader focus on a national AI framework and concerns about divergent state AI laws.
The FTC’s Deception Theory
The FTC describes AI broadly, as an umbrella term covering a range of tools and systems used across sectors. Section 5 of the FTC Act prohibits unfair or deceptive acts or practices in or affecting commerce. Under the FTC’s traditional deception framework, a representation, omission or practice is deceptive if it is likely to mislead consumers acting reasonably under the circumstances and is material to their conduct or decisions. The proposed statement applies that framework to AI systems marketed as tools that answer questions, solve problems, analyze information or assist with decision-making.
According to the FTC, consumers may reasonably expect AI systems marketed as tools for answering questions, analyzing information or assisting with decision-making to pursue objectives such as accuracy, relevance, truthfulness and responsiveness to the user’s stated goal. The Commission states that AI companies that steer outputs toward unexpected objectives, and away from objectives set by or reasonably expected by users, are likely to be viewed by the FTC as engaging in deceptive conduct under Section 5. The FTC also cites consumer reliance as part of this concern, pointing to reports that consumers accept AI system outputs without further fact-checking more than 90% of the time, even where systems are framed as striving for, rather than guaranteeing, complete accuracy.
Importantly, the FTC distinguishes intentional output steering from ordinary AI errors or “hallucinations.” The proposed statement indicates that technological limitations that lead to incorrect outputs are different from deliberate design choices that suppress accuracy or redirect outputs toward an undisclosed objective. Claims about hallucination rates or model reliability may still raise deception issues if misrepresented, but hallucinations alone are not the core conduct targeted by this proposal.
The proposal also reflects the FTC’s longstanding focus on the overall “net impression” conveyed to consumers. Under traditional FTC deception principles, the agency evaluates advertising, product design, user interfaces, disclosures and other aspects of the consumer experience collectively rather than in isolation. As a result, disclosures that technically describe output limitations or filtering practices may not be sufficient if the overall net impression created by the product, marketing or user experience suggests that the system operates differently.
State AI Regulation and Federal Preemption
The proposal devotes significant attention to the interaction between federal consumer protection law and state AI regulation. The FTC expresses concern that certain state AI frameworks could pressure AI companies to alter outputs in ways that prioritize compliance with state-defined objectives over accuracy or user-requested goals.
The Commission further states that although the FTC Act does not expressly preempt state law, state law may be preempted where it conflicts with a federal regulatory scheme. According to the FTC, “[a] State law that requires an AI firm to deceive its consumers obviously conflicts with section 5’s express purpose of protecting consumers from such conduct.”
However, the scope of the FTC’s preemption theory remains uncertain, particularly where state AI laws are framed as transparency, risk-management, accountability or anti-discrimination requirements rather than mandates to produce inaccurate outputs. Courts have traditionally been cautious about finding implied preemption absent a direct conflict between federal and state law, and the FTC’s discussion should be understood as the agency’s view rather than a definitive statement regarding the preemptive effect of Section 5.
Practical Implications for Companies Building or Using AI Tools
The proposal has practical implications for companies that build AI tools and for organizations that use them in operational, customer-facing or patient-care settings. As a starting point, companies should compare what they say about an AI tool with how the tool actually operates, including claims about accuracy, neutrality, objectivity, truthfulness, reliability, clinical or technical performance, and suitability for specific uses. Those claims may be especially important where they shape user expectations about the system’s outputs.
For companies that develop or market AI tools, the proposal underscores the importance of aligning marketing, product documentation, sales materials, user interfaces and disclosures with how the system actually behaves. Companies should consider whether disclosures adequately explain material limitations, safety guardrails, content moderation practices, output filtering, legal compliance constraints and other design choices when those practices materially affect outputs in ways that may conflict with claims or user expectations. The more a disclosure cuts against consumers’ expectations created by marketing claims or product design, the more prominent and persistent the disclosure may need to be.
For health care organizations, these issues may arise where AI tools are implemented in patient-care or clinician-facing settings. Health systems should consider how they describe those tools to clinicians, staff and patients; whether the tools are used for administrative support, patient-facing triage, care navigation, documentation, clinical decision support or diagnostic support (such as radiology or Computer-Assisted Detection/Diagnosis software); and whether training, policies, user interfaces and patient-facing materials accurately convey the tool’s intended role and limitations.
In patient-care settings, governance should also account for clinical oversight, validation for the specific use case, monitoring of output behavior, escalation pathways and vendor diligence. Health systems may want to ask vendors for information regarding intended use, validation methods, known limitations, hallucination risk, output filtering or steering practices, safety guardrails, update procedures, audit capabilities, incident reporting and material changes that may affect system outputs over time.
More broadly, companies should consider reviewing relevant marketing, product documentation, user interfaces, disclosures and governance records to assess whether statements about AI system performance align with how the system operates in practice. Contracts with AI companies should include protection for the health system in the event the AI tool engages in impermissible steering.
Looking Ahead
The proposed policy statement does not create new legal obligations, but it provides a clear signal regarding how the current FTC may evaluate AI-related deception issues. If finalized, the proposal could raise important questions regarding what constitutes “suppression of accuracy,” how the agency will distinguish impermissible output steering from legitimate safety or compliance guardrails, and when differences between user expectations and system behavior become material under Section 5.
Regardless of whether the proposal is finalized in its current form, it reinforces a theme that has appeared throughout recent FTC AI guidance: companies should be prepared to substantiate AI-related claims and ensure that representations regarding system capabilities, limitations and performance are consistent with the way the technology functions in practice.
For further information or assistance regarding this topic, please contact:
- Carolina Wirth at (202) 780-2989 or cwirth@hallrender.com;
- Melissa Markey at (248) 740-7505 or mmarkey@hallrender.com; or
- Your primary Hall Render contact.
Hall Render blog posts and articles are intended for informational purposes only. For ethical reasons, Hall Render attorneys cannot give legal advice outside of an attorney-client relationship.
