The FTC wants your AI to tell the truth (And say so if it doesn't): What the new proposed policy statement means for your business
On July 1, 2026, the Federal Trade Commission (FTC) published a proposed policy statement addressing whether AI companies violate Section 5 of the FTC Act’s prohibition on deceptive acts or practices when the companies direct their systems’ outputs toward hidden ideological objectives instead of the outcomes the consumer actually requests or reasonably expects. The public comment period closes on July 31, 2026.
Background: Executive Order 14365 and the FTC’s mandate
The proposed statement likely originates from Executive Order 14365, signed by President Trump on December 11, 2025. The executive order directed the FTC to clarify how Section 5 applies to AI models and to address potential conflicts between state laws requiring alterations to AI outputs and federal law. This reflects the administration’s larger goal of establishing a national AI regulatory framework that preempts a patchwork of state-by-state regulation.
The FTC’s position: Deception under Section 5
At its core, the FTC’s position rests on a straightforward application of the Commission’s longstanding three-party deception framework: conduct is deceptive if there is a representation, omission, or practice likely to mislead a reasonable consumer in a material way. The FTC argues that AI companies make both explicit and implicit representations that their systems are designed to produce the best, most accurate, and most faithful outputs possible within technological and resource constraints. Because consumers reportedly accept AI outputs without independent fact-checking more than 90% of the time, the Commission believes consumers reasonably expect the outputs are not secretly skewed toward undisclosed objectives.
Importantly, the proposed statement draws a clear distinction between intentional ideological steering and AI “hallucinations.” The FTC recognizes that hallucinations occur not as a result of deliberate design decisions but as a result of technological and resource limitations. Hallucinations themselves do not raise Section 5 concerns. The focus of the statement is purely on intentional decisions to train or configure a model to pursue undisclosed ideological objectives, regardless of the company’s motivation for doing so.
State law preemption: The Colorado AI Act example
Perhaps the most striking aspect of the proposed statement is the preemption analysis. The Commission specifically identifies Colorado’s Artificial Intelligence Act as an example of a state law that may pressure AI companies to suppress output accuracy to avoid disparate impact liability. The FTC concludes that such state law is impliedly preempted to the extent it conflicts with the federal regulatory scheme under Section 5. In the FTC’s view, a state law that effectively requires an AI company to deceive consumers—by forcing it to prioritize compliance objectives over accuracy without disclosure—conflicts with Section 5’s fundamental purpose of protecting consumers from deception.
This creates a potentially significant tension for companies operating in multiple jurisdictions. Compliance with certain state AI laws could, under the FTC’s theory, constitute a federal violation if the resulting output modification is not adequately disclosed to consumers.
The disclosure safe harbor: Don’t bury it
The proposed statement does, however, offer a path to compliance: an AI company can avoid Section 5 liability by making clear, conspicuous, and adequate disclosures that its system prioritizes certain objectives over what users request or expect. However, a disclaimer buried in terms of service will not be enough. The FTC expects disclosures to be prominent and persistent enough to genuinely reshape reasonable consumer expectations, which are otherwise shaped by the company’s broader marketing and by the inherent nature of AI as a problem-solving tool. Companies should not assume that existing boilerplate language will satisfy this standard.
Practical takeaways for companies
In light of the proposed statement, companies that build or use AI tools may want to take a look at the following areas:
- Align marketing claims with actual system behavior. Take stock of how your AI product is described to the public and compare that messaging against how the system is actually built and tuned. For example, is your product actually “accurate,” “unbiased,” or “objective”? Where the model has been deliberately constrained for safety, policy, or other business reasons, confirm that those constraints are made clear to users in a meaningful way, not just referenced in passing.
- Revisit how state-law obligations are being implemented. If your organization adjusts AI model outputs to satisfy a state AI statute such as Colorado’s statute, consider whether those adjustments could be construed as hiding rather than fixing a problem. This is the distinction the FTC seems to be drawing. Federal preemption may ultimately offer some cover, but that question is far from resolved, so it should not be the sole basis for a compliance strategy.
- Review existing disclosures. For any system that is designed to favor certain outcomes or objectives over a user’s actual request, take a hard look at where and how it is communicated, such as within the terms of service, product pages, and onboarding screens. Determine whether that placement would realistically catch a user’s attention and change their expectations, rather than simply checking a legal compliance box.
The public comment period closes on July 31, 2026. Companies and industry groups that wish to shape the final statement should consider submitting comments, particularly on the preemption analysis and the disclosure safe harbor standard.
McDonald Hopkins will continue to monitor this development and provide updates as the FTC finalizes its position. In the meantime, companies should begin assessing their AI systems and disclosure practices against the framework outlined in the proposed statement.
