As President Trump’s AI Action Plan and Executive Order 14365 rightly highlight, protecting American AI dominance requires curbing a chaotic, 50-state patchwork of red tape. The FTC’s proposed Policy Statement takes a commendable step by reaffirming that Section 5’s existing powers already cover deceptive AI practices without creating a legal vacuum, but the current draft falls short in key areas. To truly foster innovation, the Commission must distinguish between model developers and deployers, establish objective standards rather than trying to police subjective concepts or editorial design choices protected by the First Amendment, and create safe harbors for good-faith safety guardrails and state-law compliance. Ultimately, non-binding guidance cannot substitute for true legal preemption, so the FTC should leverage its Section 6 authority to advise Congress on a statutory national framework.
NetChoice Comments for the Record on the Federal Trade Commission Proposed Policy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems Docket No. FTC-2026-0859
July 31, 2026
NetChoice is a national trade association of leading internet and technology businesses that promotes free expression and free enterprise online. NetChoice appreciates the opportunity to submit comments to the Federal Trade Commission (FTC or the Commission) on the Proposed Policy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems (Policy Statement).
NetChoice has been strongly and publicly supportive of the Trump administration’s efforts to ensure American AI dominance. A year ago, NetChoice praised the Trump administration’s AI Action Plan for its “focus on red tape reduction” and “regulatory humility[,]” (“NetChoice Applauds President Trump’s AI Action Plan for the Golden Age of American Innovation,” NetChoice, July 23, 2025, https://netchoice.org/netchoice-applauds-president-trumps-ai-action-plan-for-the-golden-age-of-american-innovation/). Of course, red tape is not unique to the federal government. The Trump administration correctly recognized that one of the greatest risks to American AI dominance is a growing patchwork of state laws of varying aims and scopes. Subsequently, President Trump signed Executive Order 14365: Ensuring a National Policy Framework for Artificial Intelligence (Exec. Order No. 14365, 90 Fed. Reg. 58,499 (December 16, 2025), https://www.federalregister.gov/documents/2025/12/16/2025-23092/ensuring-a-national-policy-framework-for-ar tificial-intelligence). NetChoice released a statement in response to the executive order stating, “NetChoice applauds President Trump for ensuring America can lead the world in AI innovation with this executive order. As he pointed out, startups and small businesses will greatly struggle to create and compete with a 50-state patchwork of red tape. The federal government is the appropriate regulator to govern interstate commerce like AI tools[,]” (“President Trump Defends America From AI Red Tape Patchwork With New Executive Order,” NetChoice, December 11, 2025, https://netchoice.org/president-trump-defends-america-from-ai-red-tape-patchwork-with-new-executive-order/).
Among its various provisions, Executive Order 14365 instructed the FTC to “issue a policy statement on the application of the Federal Trade Commission Act’s prohibition on unfair and deceptive acts or practices under 15 U.S.C. 45 to AI models. That policy statement must explain the circumstances under which State laws that require alterations to the truthful outputs of AI models are preempted by the Federal Trade Commission Act’s prohibition on engaging in deceptive acts or practices affecting commerce” (Exec. Order No. 14365, 90 Fed. Reg. 58,499 (December 16, 2025) Sec. 7. Preemption of State Laws Mandating Deceptive Conduct in AI Models). Recognizing NetChoice’s support of President Trump’s agenda for American AI dominance and the President’s explicit instructions to the FTC to issue the proposed Policy Statement, NetChoice offers these comments in the spirit of bolstering the Trump administration’s position and consequently a competitive and thriving interstate marketplace for AI systems and services.
Emphasize Existing FTC Powers to Combat AI Harms
First and foremost, the proposed Policy Statement is most helpful in elevating the fact that AI innovation and adoption is not occurring in a legal vacuum. There is no novel technology exception to Section 5 of the FTC Act. The Policy Statement correctly emphasizes, “The Commission has applied the principles underlying its Section 5 deception authority to address false or misleading claims in a multitude of factual circumstances and across a wide variety of products and services, including new and evolving technologies” (U.S. Federal Trade Commission, Federal Trade Commission’s Proposed Policy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems (July 1, 2026), https://www.ftc.gov/system/files/ftc_gov/pdf/ai-policy-statement_0.pdf). A subsequent final Policy Statement should continue to hammer home this point. It is a powerful reminder to state legislators and their constituents that there is already applicable law and a cop on the beat when it comes to potential deceptive or fraudulent acts covering AI systems and their misuse. NetChoice encourages the FTC to make this point unambiguously in the final Policy Statement and continuously in public remarks and statements. Such will deter state legislatures from layering on superfluous rules and regimes regarding one of the more prominent consumer concerns about AI.
Address the Developer/Deployer Distinction
Executive Order 14365 asked the Commission to address the application of Section 5 to “AI models” (Exec. Order No. 14365, 90 Fed. Reg. 58,499 (December 16, 2025)). The proposed Policy Statement addresses itself throughout to “AI companies.” That phrase is broad enough to cover both a startup firm shipping a customer service assistant (a deployer) and the laboratory that trained the model running underneath it (the developer). These are not the same business. They do not stand in the same relationship to the consumer, and they cannot discharge the same obligations. The “companies” for “models” substitution silently widens the field of covered parties from developers to include deployers, and a final Policy Statement should not leave the resulting ambiguity for enforcement to resolve.
The startup building the customer service assistant selects a model, tunes it, writes the instructions that frame every exchange, decides what the product will refuse, designs the screen, drafts the marketing copy, and sets the terms of service. The laboratory that trained the underlying model may never learn that the application exists. It cannot see the interface, cannot audit the configuration, and cannot place a word in front of that application’s users. Holding the lab to a disclosure duty it has no mechanism to perform is not consumer protection but a penalty untethered from any capacity to prevent the harm.
NetChoice accordingly asks the Commission to make four things explicit in a final Policy Statement.
- First, the party that puts the product in front of the public, and makes the representations that shape what the public expects, is the party that owes the public a disclosure.
- Second, the technical documentation that model providers already publish for the businesses licensing their models is the appropriate and sufficient channel for communicating design choices and limitations upstream of the consumer.
- Third, no firm should face Section 5 exposure for output-shaping decisions made by a counterparty, outside its knowledge and beyond its control.
- Fourth, the reasonable consumer standard should be measured against the audience actually addressed; where a provider’s counterparty is a sophisticated commercial licensee rather than a member of the public, the deception analysis is different in kind, and the Commission should say so.
Prioritize Objective Standards over Subjective State Regulations
All questions and tensions in public policy at some level boil down to a matter of objectivity versus subjectivity. Regulation is most effective when it focuses on the former. Objective standards create the necessary predictability for the kind of intensive investment and innovation the Trump administration seeks to foster in the AI industry.
An impetus for this proposed Policy Statement is the fact that the State of Colorado opened the door to the subjective value judgments of state officials in influencing the development of AI systems. As NetChoice member xAI stated in its lawsuit against Colorado’s SB 24-205, “[T]he bill defines ‘[a]lgorithmic discrimination’ as ‘any condition in which the use of an [AI] system results in unlawful differential treatment or impact that disfavors an individual or group of individuals on the basis of’ certain protected characteristics. § 6-1-1701(1)(a). But not all ‘algorithmic discrimination’ is prohibited. The bill proscribes only discrimination that Colorado disagrees with. Indeed, it expressly exempts from its definition of ‘algorithmic discrimination’ any discrimination that ‘[e]xpand[s] an applicant, customer, or participant pool to increase diversity or redress historical discrimination,’ § 6-1-1701(1)(b)(I)(B), thereby importing Colorado’s normative judgment of what preferences should be given to certain groups” (Complaint for Declaratory and Injunctive Relief, X.AI LLC v. Weiser, No. 1:26-cv-01515 (D. Colo. filed Apr. 9, 2026), ECF No. 1. https://storage.courtlistener.com/recap/gov.uscourts.cod.253513/gov.uscourts.cod.253513.1.0.pdf).
While 50 different sets of rules governing a particular technology or industry is burdensome enough, that burden becomes effectively insurmountable if state officials, such as attorneys general, with diametrically opposing political and cultural views are enabled to enforce those regimes based on subjective versus objective criteria. Compliance in one jurisdiction could easily trigger liability in another. No amount of resources dedicated to compliance or trust and safety resolves this issue. The only way to ensure no liability is to not participate in the market. NetChoice cautions the FTC to not cause the federal government to simply become a 51st jurisdiction of competing interests. To this end, NetChoice advises the following:
Safety and integrity design choices must be carved out. NetChoice appreciates the Commission’s recognition of consumer expectations as a baseline for Section 5 analysis, but urges the FTC to clarify that good-faith safety guardrails and harm-prevention design choices fall entirely outside the scope of output “deception.” When a user interacts with an AI product, even one marketed as accurate, neutral, or objective, the reasonable consumer expectation is that the system will decline to facilitate illegal, dangerous, or harmful acts. Refusing to provide instructions for illegal activities or steering away from high-risk content is not a deceptive deviation from user expectations. Consumers expect as much from responsible services.
Under the current draft, companies face significant disclosure exposure for routine product-safety engineering. A product held to a standard of absolute accuracy risks Section 5 scrutiny whenever its outputs are shaped by undisclosed design choices, even when the sole motivation is safety, harm classification, or legal compliance. This dynamic creates a paradox: while the statement properly carves out unintentional hallucinations from Section 5 liability, it inadvertently subjects deliberate safety interventions to heightened enforcement risk. The logic applied to inherent technical limitations like hallucinations compellingly applies to safety refusals as well. If unassisted, probabilistic errors are not deceptive, then intentional, good-faith guardrails against severe risks and objective harms like child exploitation, dangerous materials, or cyberattacks cannot logically be treated as covert “suppression of accuracy.”
To resolve this tension and prevent a chilling effect on responsible AI development, the Commission should expand its proposed cybersecurity carve-out into a generalized safe harbor for good-faith harm prevention. The FTC should affirm that safety, integrity, and compliance guardrails fall outside the deception analysis entirely. Consumers do not expect AI systems to act as unrestricted engines for unlawful conduct. Declining to assist with harm is not a suppression of truth, and the FTC should reset its framework to ensure companies are not penalized for engineering safe, law-abiding products.
Accuracy isn’t a coherent enforceable standard. NetChoice recognizes the FTC has been put in a precarious position where if precision is not exercised the Commission will find itself doing the same thing this proposed Policy Statement attempts to prevent state governments from doing. From a practical standpoint, NetChoice urges the Commission to recognize that “accuracy” cannot be applied as a blunt, binary standard across artificial intelligence outputs. Because large language models are inherently probabilistic rather than deterministic systems, treating perfection or absolute correctness as a baseline fundamentally misapprehends how generative AI functions. Furthermore, AI outputs routinely involve nuance, synthesis, and prioritization across subjective or contested domains. Declaring an output “inaccurate” or “deceptive” overlooks the reality that many queries do not yield a single objective answer, but rather require the system to weigh complex information and exercise context-dependent judgment.
To establish a workable framework that avoids penalizing standard model architecture, the FTC must scope its accuracy expectations strictly to outputs designed and presented to convey factual information, and only where a consumer would reasonably expect as much. AI systems are deployed across a vast spectrum of use cases, from storytelling to matters of open debate, where concepts of “truthfulness” or “accuracy” are wholly inapplicable. Consider, for example, the fact that the practice of law itself involves practitioners, lawyers and judges alike, arriving at differing conclusions based on the same set of facts. Or consider scientific study. There is an obvious difference between the questions “Which atoms form a water molecule?” versus “How did the universe begin?” Even the question, “What is science?” arguably lacks a perfectly accurate answer. “[T]he question of the scientific method is one of the most difficult, most contentious, most puzzling problems in modern thought,” wrote Michael Strevens in The Knowledge Machine. Strevens continues that in practice, “scientists seem scarcely to follow any rules at all” (Michael Strevens, The Knowledge Machine: How Irrationality Created Modern Science (New York: Liveright, 2020)).
The FTC must avoid policing complicated topics or theories as well as moral and political questions which inherently lack “factual” resolutions (See, e.g., Richard A. Posner, “The Problematics of Moral and Legal Theory,” 111 Harvard Law Review 1637 (1998) (“You cannot show me that my [moral] intuition is an illusion, like the apparent movement of the sun or the bent appearance of a straight stick in water. There are no ‘crucial experiments,’ and no statistical regularities, by which to validate a moral argument.”)). The Commission should explicitly clarify that Section 5 scrutiny applies only where an AI system purports to deliver factual claims and a consumer would reasonably rely on those outputs as objective truth while acknowledging that for many matters subjective decisions will have to be made regarding the information provided. Doing so is not only practically prudent but more legally durable.
Constitutional Limitations
The FTC must considerably narrow its approach on content not only as a practical matter but as a constitutional one. As established in Moody v. NetChoice (Moody v. NetChoice, LLC, 603 U.S. 707 (2024) (decided together with NetChoice, LLC v. Paxton, No. 22-555) https://www.supremecourt.gov/opinions/23pdf/22-277_d18f.pdf), processes of ranking, filtering, and refining expressive outputs are protected editorial activities. Government attempts to mandate or restrict these content-based design decisions must meet heightened First Amendment scrutiny. These First Amendment protections apply to the FTC as strictly as they do to the states. Any enforcement mechanism that judges a firm’s “ideological objectives” in model outputs inevitably requires the government to oversee and penalize editorial choices. This runs headlong into the First Amendment. Users have the expectation and the right to receive information free from government interference or manipulation whether it comes from state law or a federal agency.
To further construct a legally resilient framework, the FTC must anchor its preemption arguments in established procedural protocols and constitutional frameworks, rather than relying on mere agency guidance. The proposed Policy Statement identifies itself as a general policy declaration, outlining the Commission’s planned exercise of Section 5 enforcement discretion without carrying the force of law. Because it was not adopted through notice-and-comment rulemaking, it lacks the authority to bind the Commission, the judiciary, or market participants. This distinction is critical: federal preemption necessitates the supremacy of actual law.
Furthermore, the Commission cannot manufacture legal authority by simple decree. In the absence of a specific congressional delegation of preemptive power, an agency’s claim that it has displaced state law is granted no special deference; its validity depends entirely on the inherent persuasiveness of its reasoning. Section 5 contains no express preemption provision. Following the Supreme Court’s ruling in Loper Bright Enterprises v. Raimondo (Loper Bright Enterprises v. Raimondo, 603 U.S. 369 (2024) https://www.supremecourt.gov/opinions/23pdf/22-451_7m58.pdf), courts will evaluate the Commission’s jurisdictional reach independently, without deferring to the agency’s own interpretation of its power.
Practically, a firm relying on this Policy Statement as protection against state-level enforcement would find itself without a viable defense. State attorneys general and state courts remain unswayed by the Commission’s non-binding views. Consequently, businesses could find themselves in a trap: facing federal scrutiny for conduct that states require, while suffering state liability that federal guidance is powerless to prevent.
The weakness of this implied preemption theory is further highlighted by the Executive Branch’s own legal strategy. When contesting Colorado’s previous AI law alongside xAI, the Department of Justice avoided preemption arguments, opting instead for constitutional challenges under the First and Fourteenth Amendments (United States’ Complaint in Intervention, X.AI LLC v. Weiser, No. 1:26-cv-01515-DDD-CYC (D. Colo. filed Apr. 24, 2026), ECF No. 12-2. https://www.justice.gov/crt/media/1437846/dl).
Utilize the Commission’s Advisory Role to Assist Congress
Despite the constitutional limitations around the proposed Policy Statement, the FTC certainly has a vital role to play. Rather than stretching Section 5 enforcement powers, the Commission should leverage its Section 6 investigative and reporting authority. This fact-finding mission was the original purpose for which the Commission was established.
The FTC was designed to provide Congress with expert, nonpartisan analysis of commercial practices that case-by-case litigation could not offer. Section 6 empowers the Commission to compile information on corporate conduct, demand special reports, and provide findings to Congress alongside “recommendations for additional legislation” (15 U.S.C. § 46(f)). This advisory function is a core part of its architecture, not a secondary feature.
NetChoice suggests that the Commission complement the Policy Statement with a Section 6(f) report to Congress. Such a report should specifically identify conflicting state mandates, document the engineering and compliance burdens on developers, and propose a substantive federal standard. Crucially, it should recommend an express preemption provision to clarify which state laws are displaced. This would provide Congress with a coherent, fact-based framework and produce a uniform national standard that is immune to the constitutional and procedural vulnerabilities of the current proposal.
Establish Interim Safe Harbors to Protect Innovation
Since the legislative process is gradual, the FTC must ensure the interim period does not become a trap for innovators. The Commission’s application of the “reasonable consumer” standard creates a logical contradiction. Section 5 deception requires a practice likely to mislead a reasonable consumer regarding a material fact, yet “reasonableness” must account for existing law. When state legislatures adopt and publicly enforce output requirements, a consumer in that state logically expects companies to follow those laws. Compliance with transparent legal mandates aligns with consumer expectations rather than subverting them. Treating adherence to an active state statute as “deceptive” would effectively penalize firms for following the law, which is an untenable standard. To avoid a situation where firms face impossible choices between state duties and federal enforcement, the Commission should implement clear safe harbors to guide compliant behavior.
An affirmative safe harbor should protect companies from Section 5 liability when they are acting in good faith to comply with active state laws. This protection should be based on objective markers, such as the existence of a state requirement and clear disclosure at the point of user access, rather than requiring expensive litigation to prove intent.
The Commission should also establish a formal notification mechanism, such as a referral process with the Department of Justice’s AI Litigation Task Force under Executive Order 14365, to identify conflicting statutes in advance. Providing clarity through public lists rather than through enforcement actions would protect smaller developers from the disproportionate costs of serving as legal test cases. Finally, the Commission must explicitly disclaim any retroactive enforcement.
By combining these interim safeguards with a Section 6(f) recommendation, the Commission can help achieve what a mere policy statement cannot: a stable, constitutionally sound environment for American AI dominance where obligations are clear, state authority is respected where appropriate, and consumers enjoy meaningful but unintrusive protections.
NetChoice appreciates the FTC’s work to advance the Trump administration’s pursuit of American AI dominance. We recognize that the Commission is under explicit instruction from President Trump to produce a Policy Statement on its Section 5 authority in regards to state AI laws. Any final Policy Statement will not be a silver bullet in preventing an onerous and ideological patchwork of state laws around AI development. However that does not mean a final Policy Statement and further FTC efforts cannot be meaningful contributions to establishing a necessary federal policy framework on AI.
Sincerely,
Patrick Hedger
Director of Policy, NetChoice (The views of NetChoice expressed here do not necessarily represent the views of all NetChoice members.)
NetChoice is a trade association that works to protect free expression and promote free enterprise online.