{"slug": "policy-statement-concerning-the-suppression-of-accuracy-in-ai-systems", "title": "Policy Statement Concerning the Suppression of Accuracy in AI Systems", "summary": "The Federal Trade Commission issued a policy statement concerning the suppression of accuracy in artificial intelligence systems, seeking public comments until July 31, 2026. The statement addresses regulatory actions to ensure AI accuracy and prevent deceptive practices.", "body_md": "This site displays a prototype of a “Web 2.0” version of the daily\nFederal Register. It is not an official legal edition of the Federal\nRegister, and does not replace the official print version or the official\nelectronic version on GPO’s govinfo.gov.\n\nThe documents posted on this site are XML renditions of published Federal\nRegister documents. Each document posted on the site includes a link to the\ncorresponding official PDF file on govinfo.gov. This prototype edition of the\ndaily Federal Register on FederalRegister.gov will remain an unofficial\ninformational resource until the Administrative Committee of the Federal\nRegister (ACFR) issues a regulation granting it official legal status.\nFor complete information about, and access to, our official publications\nand services, go to\nAbout the Federal Register\non NARA's archives.gov.\n\nThe OFR/GPO partnership is committed to presenting accurate and reliable\nregulatory information on FederalRegister.gov with the objective of\nestablishing the XML-based Federal Register as an ACFR-sanctioned\npublication in the future. While every effort has been made to ensure that\nthe material on FederalRegister.gov is accurately displayed, consistent with\nthe official SGML-based PDF version on govinfo.gov, those relying on it for\nlegal research should verify their results against an official edition of\nthe Federal Register. Until the ACFR grants it official status, the XML\nrendition of the daily Federal Register on FederalRegister.gov does not\nprovide legal notice to the public or judicial notice to the courts.\n\nNotice\n\nEnter a search term or FR citation e.g.\n88 FR 38230 FR 78782024-13208USDA09/05/24RULE0503-AA39SORN\n\nChoosing an item from\nfull text search results\nwill bring you to those results. Pressing enter in the search box\nwill also bring you to search results.\nChoosing an item from\nsuggestions\nwill bring you directly to the content.\n\nThank you for taking the time to create a comment. Your input is important.\n\nOnce you have filled in the required fields below you can preview and/or submit your comment to the Federal Trade Commission for review. All comments are considered public and will be posted online once the Federal Trade Commission has reviewed them.\n\nComments must be received on or before Friday, July 31, 2026.\n\nTable of Contents\n\nEnhanced Content - Table of Contents\n\nThis table of contents is a navigational tool, processed from the\nheadings within the legal text of Federal Register documents.\nThis repetition of headings to form internal navigation links\nhas no substantive legal effect.\n\nAgencies review all submissions and may choose to redact, or withhold,\ncertain submissions (or portions thereof). Submitted comments may not be\navailable to be read until the agency has approved them.\n\nDocket Title\n\nDocument ID\n\nComments\n\nPolicy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems\n\nFederalRegister.gov retrieves relevant information about this document\nfrom Regulations.gov to provide users with additional context. This\ninformation is not part of the official Federal Register document.\n\nPolicy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems\n\nDocument page views are updated periodically throughout the day and are\ncumulative counts for this document. Counts are subject to sampling,\nreprocessing and revision (up or down) throughout the day.\n\nPage views\n\n404\nas of\n07/08/2026 at 6:15 pm EDT\n\nOther Formats\n\nEnhanced Content - Other Formats\n\nThis document is also available in the following formats:\n\nThis PDF is FR Doc. 2026-13628 as it appeared on Public Inspection on\n07/06/2026 at 8:45 am.\n\nIt was viewed\n123\ntimes while on Public Inspection.\n\nIf you are using public inspection listings for legal research, you\nshould verify the contents of the documents against a final, official\nedition of the Federal Register. Only official editions of the\nFederal Register provide legal notice of publication to the public and judicial notice\nto the courts under 44 U.S.C. 1503 & 1507.\nLearn more here.\n\nPublished Document: 2026-13628 (91 FR 41638)\n\nThis document has been published in the Federal Register. Use the PDF linked in the document sidebar for the official electronic format.\n\nAGENCY:\n\nFederal Trade Commission.\n\nACTION:\n\nProposed policy statement; request for comments.\n\nSUMMARY:\n\nThe Federal Trade Commission (“Commission”) is proposing a policy statement regarding the application of the prohibition on deceptive acts or practices in section 5 of the Federal Trade Commission Act to companies that market artificial intelligence (“AI”) systems.\n\nDATES:\n\nComments must be received on or before Friday, July 31, 2026.\n\nADDRESSES:\n\nMembers of the public may file a comment online or on paper by following the instructions in the Comment Submissions part of the\nSUPPLEMENTARY INFORMATION\nsection below. Write “AI Policy Statement; Matter No. P264200” on your comment and file your comment online at\nhttps://www.regulations.gov/docket/FTC-2026-0859.\nIf you prefer to file your comment on paper, mail your comment to the following address: Federal Trade Commission, Office of the Secretary, 600 Pennsylvania Avenue NW, Mail Stop H-144 (Annex P), Washington, DC 20580.\n\nSUPPLEMENTARY INFORMATION:\n\nI. Summary\n\nArtificial Intelligence (AI) is reshaping how Americans consume information, educate our children, and perform our jobs.[1]\nAs they have marketed their remarkable breakthroughs to the public, AI companies have spent years representing explicitly and implicitly that their systems aim to produce the best output—output that faithfully and accurately achieves users' stated objectives and the built-in objectives that users expect in the AI system—that is possible within their technological and resource constraints. Because of these representations and the inherent nature of the products and services in question, consumers have a reasonable expectation that AI systems aim to give truthful and accurate outputs. Consumers have no basis to believe that AI systems aim to produce outputs that are distorted by undisclosed ideological objectives.\n\nNonetheless, an AI company might be tempted to alter or steer the output of its systems contrary to consumers' reasonable expectations for various reasons, including attempted compliance with a State law, such as Colorado's recently revised Artificial Intelligence Act. But steering an AI system in this manner may deceive consumers in violation of section 5 of the FTC Act. That is true even if the deceptive steering is done in an effort to comply with State laws. Of course, a company may be able to avert potential deception by making truthful, non-misleading representations about the aims of its model. But such representations would need to make clear the AI company is prioritizing objectives different than those consumers requested or would otherwise expect.\n\nII. Background\n\nArtificial intelligence is an umbrella term covering a universe of different tools and systems now used across nearly every sector of the economy and in most people's daily lives.[2]\nRegardless of whether one has in mind modern large language models, AI applications applying large-language models to particular purposes, or a not-yet-developed superintelligence, the hallmark of a successful AI system is an ability to accurately “make predictions, recommendations, or decisions” for its user consistent with a given objective.[3]\nTheir utility can then be judged by how well the solution matches consumers' objectives.\n\nOn a fundamental level, these products and services solve problems for people. For many Americans, they are becoming part and parcel of daily life, relied upon for research, analysis, and advice on both personal and business issues. This development reflects consumers' confidence that AI companies are bound by the same rules as other American companies: they will deal with consumers in good faith and with no hidden agenda, and they will not work in the shadows to punish those who hold opinions contrary to theirs. In large part, that confidence itself reflects the basic fact that the United States is the global standard-bearer for AI technologies.[4]\nThat American companies dominate every layer of the AI ecosystem should not be taken for granted. Geopolitical rivals are investing heavily in this sphere, hoping to inject their own companies and values into the marketplace. American dominance is thus about more than winning some abstract race. It is about ensuring Americans can continue to feel their values are being respected as they interact with, and benefit from, an AI-powered economy.\n\nUnder the President's leadership, the Trump-Vance Administration has taken decisive steps to sustain America's global AI dominance [5]\nby removing regulatory barriers that impede AI innovation.[6]\nExcessive AI regulation would undermine American AI supremacy by deterring and suppressing the same ingenuity responsible for making American AI great.[7]\nAt the same time, however, as President Trump's National Policy Framework for Artificial Intelligence recognizes, responsible AI innovation can co-exist with prudent guardrails. For example, AI technologies should protect children and empower\n( printed page 41639)\nparents.[8]\nAnd most relevant to this statement, AI technologies should not be used to silence or censor lawful expression or dissent.[9]\nImportantly, President Trump's proposed approach is a national AI framework, protecting innovation and competition by providing national regulatory clarity and certainty and avoiding a balkanized or patchwork regulatory approach driven by the States—or, most dangerously, imposed by certain anti-innovation State governments on the rest of the country.\n\nAmerica's AI Action Plan and other critical executive actions strike that balance between accelerating innovation and protecting Americans,[10]\nbut anti-innovation States' recent efforts to regulate AI are concerning. The growing number of enacted and proposed State AI laws threatens to create a patchwork of regulatory regimes and compliance challenges for American companies.[11]\nFor example, some States have enacted laws regulating the outputs of various AI models, ultimately requiring American companies to embed “ideological bias within [their AI] models.” [12]\n\nOn December 11, 2025, President Trump signed Executive Order 14365, “Ensuring a National Policy Framework for Artificial Intelligence.” [13]\nThe Executive Order builds on the Administration's prior work to encourage adoption of AI and remove regulatory barriers, aiming to “sustain and enhance the United States' global AI dominance” by establishing a “minimally burdensome national policy framework for AI—not 50 discordant State ones.” [14]\nThe Executive Order directs the Commission to issue this enforcement policy statement clarifying the application of section 5 of the Federal Trade Commission Act (“FTC Act”) to AI models, and in particular, address how State laws requiring alterations to the accurate outputs of AI models can conflict with the requirements of the FTC Act.[15]\n\nIII. The Commission's Authority Under Section 5 of the FTC Act\n\nNearly ninety years ago, Congress gave the Commission the authority to protect consumers from “unfair or deceptive acts or practices in or affecting commerce.” [16]\nImportantly, Congress did not provide State-law safe harbors—even where a State's laws shape the provision of those products or services, companies must comply with section 5.\n\nSection 5 of the FTC Act's prohibition on deceptive acts or practices secures consumers' right to truth in the marketplace.[17]\nIn our 1983 Policy Statement on Deception, we said “the Commission will find deception if there is a representation, omission or practice that is likely to mislead the consumer acting reasonably in the circumstances, to the consumer's detriment.” [18]\nThis test, which “undergird[s] all deception cases,” consists of three elements.[19]\n“\nFirst,\nthere must be a representation, omission or practice that is likely to mislead the consumer.” [20]\n“\nSecond,\nwe examine the practice from the perspective of a consumer acting reasonably in the circumstances.” [21]\nAnd “[\nt]\nhird,\nthe representation, omission or practice must be a material one,” such that it is “likely to affect the consumer's conduct or decision with regard to a product or service.” [22]\n\nFederal courts have adopted the same principles in assessing whether an act or practice is deceptive.[23]\nCourts have also held the Commission's deception authority covers implied misrepresentations, half-truths,[24]\nand instances where only a “significant minority” of consumers are misled.[25]\nMoreover, courts have recognized the Commission's deception authority requires the entity to substantiate the representation it makes about its products or services with sufficient evidence.[26]\nAt its core, the Commission's deception authority is based on “well-established principles of advertising law and common sense.” [27]\n\nThe 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,[28]\nincluding new and evolving technologies.[29]\nAnd we have taken aim at companies that failed to provide adequate disclosures,[30]\nas well as\n( printed page 41640)\ngovernmental and private standard-setting entities that sought to restrict truthful advertising.[31]\n\nIV. Deceiving Consumers as to the Objectives of an AI System Is a Section 5 Violation\n\nAI products and services are not exempt from section 5's long-established, and generally applicable reach. The Commission has already brought a number of recent enforcement actions to combat deceptive claims in connection with AI-powered scams, as well as material misrepresentations about the performance, efficacy, and characteristics of AI products.[32]\nFor instance, the Commission recently invoked well-established section 5 principles to address an AI company's misleading claims about the ability of its conversational AI tool to replace human customer service representatives.[33]\nAlthough the Commission is careful to avoid unduly burdening innovation in the AI industry,[34]\nthe Commission will continue to protect consumers and the market from deceptive or unfair business practices, including those relying on AI systems or technologies, by using well-established section 5 principles.\n\nIn marketing their products as problem-solving tools, AI companies have represented explicitly and implicitly that their AI systems aim to produce the best output possible given technological and resource constraints.[35]\nAI companies aggressively market themselves as building products and services that distill all of human knowledge to solve problems, large and small, in furtherance of users' given objectives.[36]\nThis is because companies know their success depends on their products' utility, and an AI system's utility is a function of its ability to accomplish the objectives users ask it to accomplish.[37]\n\nConsumers share this understanding of the purpose of AI systems and have learned to trust these systems across a broad swath of applications, from education to health, finance, relationships or a myriad of other uses.[38]\nIn fact, according to one major AI developer, consumers accept AI system outputs without conducting any further fact checking over 90% of the time, even though these systems purport to only strive for, rather than guarantee, complete accuracy.[39]\nIn other words, consumers ask AI systems for help answering deeply personal and important questions, and they overwhelmingly trust that the system is designed to provide an answer tailored specifically to their objectives.[40]\nAnd because of both the companies' marketing and the inherent value proposition of an AI system, this widely\n( printed page 41641)\nheld consumer expectation is eminently reasonable.\n\nOf course, AI systems are likely to simultaneously pursue multiple objectives, some explicit and some implicit. This will often be consistent with consumers' reasonable expectations. Users of an AI chatbot might, for example, reasonably expect the system to balance succinctness, clarity, relevance, accuracy, and other objectives in its attempt to produce the best output. And while truth and accuracy are in many cases implicit objectives in requests to AI systems, a user could request output that is intentionally inaccurate or that deprioritizes accuracy in favor of some other objective like entertainment.\n\nThese representations and baseline consumer expectations notwithstanding, AI companies might steer their systems' outputs toward objectives other than those that consumers ask for or expect. A company could be tempted, for example, to abuse consumer trust by training a model surreptitiously to produce ideologically motivated distortions in a response to a factual question, such as to correct what the developer believes are “historical injustices” in the facts.[41]\nOr a company could be pressured by a State law to alter its technology's outputs. The original version of Colorado's Artificial Intelligence Act,[42]\nfor example, imposed a broad duty on AI companies to avoid output that might lead to disparate impacts in various contexts, including when a customer's foreseeable use of that output could itself create a disparate impact. And the revised version of that law explicitly provides AI companies can be held liable for discriminatory outcomes caused by their customers' use of their products.[43]\nIt is predictable that an AI company might suppress accuracy and interpose other objectives, such as so-called “equity,” to avoid liability under this law, but fail to disclose these ulterior objectives in order to hide the loss of accuracy they necessitate.[44]\nAI companies may also be subject to public pressure to surreptitiously avoid politically inflammatory outputs, as well as the temptation to clandestinely modify their AI systems to further their own or their employees' political agendas.\n\nIn light of the above, the Commission believes AI companies that steer the outputs of their AI systems toward unexpected objectives, and away from the objectives set by or reasonably expected by users, are likely to deceive [45]\nconsumers in violation of section 5 of the FTC Act.[46]\nAI companies have made both explicit and implicit representations that their systems aim to produce outputs that achieve users' objectives as faithfully and accurately as they can.[47]\nThose representations, absent adequate disclaimers or qualifications, are material as consumers rely on them and are likely to consider, in their decision as to which AI system to use or pay for, whether a particular system is designed to produce output accomplishing their objectives or to intentionally produce a worse output in service of another objective. Consumers may be induced to pay for a service that does not behave as advertised. Consumers may also be deceived into relying on a technology that, by design, produces worse outputs and may recommend suboptimal courses of action, not because of any technological or resource limitations, but because the AI developer's hidden agenda subverted consumers' objectives.\n\nAs is always the case, a company's motives for deceiving consumers are irrelevant to the application of section 5.[48]\nWhether motivated by profit, shaping public opinion, or anything else, section 5 prohibits deceiving consumers. These prohibitions apply even when a company engages in a deceptive act or practice in order to comply with a State law. Although the FTC Act does not expressly preempt State law, State law is impliedly preempted to the extent it conflicts with a Federal regulatory scheme.[49]\nA 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.\n\nAn AI company can, of course, take actions that shape consumer expectations in ways that make it unreasonable for consumers to believe the AI system is designed to achieve the objectives that consumers would otherwise expect. A company can, for example, clearly and conspicuously disclose that its systems are designed to produce outputs that prioritize certain objectives over what users request and otherwise expect. But such a disclaimer would have to be adequate to shift consumer expectations that would be based otherwise both on companies' explicit representations and on the inherent value proposition of AI systems as tools to solve human problems. An adequate disclaimer could not be buried in terms of service, for instance. It would have to clearly and conspicuously dispel the notion that the system is designed to give the best answer possible. Such a disclaimer would need to be prominent, and it is doubtful a one-time disclosure subsequently hidden away in fine print would suffice. The more the disclosure cuts against the reasonable expectations users would take away from other contexts, the more persistent and prominent it would need to be.[50]\nA prominent misrepresentation is unlikely to be remedied by a less prominent,\n( printed page 41642)\nsubsequent disclosure.[51]\nUltimately, it is the company's responsibility to ensure compliance with section 5.\n\nV. Comment Submissions\n\nYou can file a comment online or on paper. For the Commission to consider your comment, we must receive it on or before Friday, July 31, 2026. Write “AI Policy Statement; Matter No. P264200” on your comment. Your comment—including your name and your State—will be placed on the public record of this proceeding, including, to the extent practicable, on the\nhttps://www.regulations.gov\nwebsite.\n\nWe encourage you to submit comments through the\nhttps://www.regulations.gov\nwebsite. Postal mail addressed to the Commission will be subject to delay because of heightened security screening. If you prefer to file your comment on paper, write “AI Policy Statement; Matter No. P264200” on your comment and on the envelope, and send it via overnight service to: Federal Trade Commission, Office of the Secretary, 600 Pennsylvania Avenue NW, Mail Stop H-144 (Annex P), Washington, DC 20580.\n\nBecause your comment will be placed on the publicly accessible website at\nhttps://www.regulations.gov,\nyou are solely responsible for making sure your comment does not include any sensitive or confidential information. In particular, your comment should not include sensitive personal information, such as your or anyone else's Social Security number; date of birth; driver's license number or other State identification number, or foreign country equivalent; passport number; financial account number; or credit or debit card number. You are also solely responsible for making sure your comment does not include sensitive health information, such as medical records or other individually identifiable health information. In addition, your comment should not include any “trade secret or any commercial or financial information which . . . is privileged or confidential”—as provided by section 6(f) of the FTC Act, 15 U.S.C. 46(f), and FTC Rule 4.10(a)(2), 16 CFR 4.10(a)(2)—including competitively sensitive information such as costs, sales statistics, inventories, formulas, patterns, devices, manufacturing processes, or customer names.\n\nComments containing material for which confidential treatment is requested must be filed in paper form, must be clearly labeled “Confidential,” and must comply with FTC Rule 4.9(c). In particular, the written request for confidential treatment that accompanies the comment must include the factual and legal basis for the request and must identify the specific portions of the comment to be withheld from the public record.\nSee\nFTC Rule 4.9(c). Your comment will be kept confidential only if the General Counsel grants your request in accordance with the law and the public interest. Once your comment has been posted on the\nhttps://www.regulations.gov\nwebsite—as legally required by FTC Rule 4.9(b)—we cannot redact or remove your comment from that website, unless you submit a confidentiality request that meets the requirements for such treatment under FTC Rule 4.9(c), and the General Counsel grants that request.\n\nThe FTC Act and other laws the Commission administers permit the collection of public comments to consider and use in this proceeding, as appropriate. The Commission will consider all timely and responsive public comments it receives on or before Friday, July 31, 2026. For information on the Commission's privacy policy, including routine uses permitted by the Privacy Act, see\nhttps://www.ftc.gov/site-information/privacy-policy.\n\n2.\nSee,\ne.g.,E.O. 14319, Preventing Woke AI in the Federal Government, at sec. 1, 90 FR 35389, 35389 (July 23, 2025) (“Artificial intelligence . . . will play a critical role in how Americans of all ages learn new skills, consume information, and navigate their daily lives.”).\n\n5.\nSee,\ne.g.,\nAI Action Plan,\nsupra\nnote 1; E.O. 14318, Accelerating Federal Permitting of Data Center Infrastructure, 90 FR 35385, 35385 (July 23, 2025) (prioritizing the rapid and efficient buildout of AI data centers and the infrastructure that power them by easing Federal regulatory burdens); E.O. 14319, 90 FR at 35389 (noting agencies have an “obligation not to procure [AI] models that sacrifice truthfulness and accuracy to ideological agendas”); E.O. 14320, Promoting the Export of the American AI Technology Stack, 90 FR 35393, 35393 (July 23, 2025) (establishing a coordinated national effort to support the American AI industry by promoting the export of full-stack American AI technology packages).\n\n12.\nSee,\ne.g.,E.O. 14365, Ensuring a National Policy Framework for Artificial Intelligence, 90 FR 58499, 58499 (Dec. 11, 2025) (“For example, a new Colorado law banning `algorithmic discrimination' may even force AI models to produce false results in order to avoid a `differential treatment or impact' on protected groups.”); AI National Policy Fact Sheet,\nsupra\nnote 11 (“States such as California and Colorado are considering requiring AI companies to censor outputs and insert left-wing ideology in their programming.”). Colorado has since materially revised the law referred to by this Executive Order, but the new version poses many of the same concerns. See\ninfra\nnote 43.\n\n22.\nId.\nMaterial information affects a consumer's decision to purchase a product but may also affect other kinds of consumer conduct.\nId.\nat 182 n.45 (citing the Commission's complaint in\nVolkswagen of America,\n99 F.T.C. 446 (1982)).\n\n27.\nFTC\nv.\nCom. Planet, Inc.,\n878 F. Supp. 2d 1048, 1083 (C.D. Cal. 2012),\naff'd in part,\n642 F. App'x 680 (9th Cir. 2016); see also\nFTC\nv.\nJohnson,\n96 F. Supp. 3d 1110, 1142 (D. Nev. 2015) (rejecting “fair notice” argument as to section 5(a) liability and noting there is “extensive case law and guidance on what constitutes a deceptive act or practice under Section 5(a)”).\n\n28.\nSee,\ne.g., FTC\nv.\nFleetcor Techs., Inc.,\n620 F. Supp. 3d 1268, 1289-1313 (N.D. Ga. 2022) (fuel cards and discounts),\naff'd sub nom FTC\nv.\nCorpay, Inc.,\n164 F.4th 807 (11th Cir. 2026);\nFTC\nv.\nD-Link Sys., Inc.,\nNo. 3:17-CV-00039-JD, 2017 WL 4150873, at *2 (N.D. Cal. Sept. 19, 2017) (data security and protections for IoT devices);\nPom Wonderful,\n777 F.3d at 490-500 (dietary supplements);\nCom. Planet, Inc.,\n878 F. Supp. 2d at 1083 (online web pages and negative option plans).\n\n29.\nSee,\ne.g.,\nStipulated Order,\nFTC\nv.\nAmazon.com, Inc.,\nNo. 2:23-CV-00932-JHC (W.D. Wash. Sept. 25, 2025), Dkt. No. 535 (enrolling consumers into online shopping memberships without consumers' express informed consent); Compl.,\nUnited States\nv.\nFacebook,\nNo. 19-cv-2184 (D.D.C. July 24, 2019), Dkt. No. 1 (misrepresenting that users would have to “turn[] on” facial-recognition technology); Compl.,\nFTC\nv.\nNew Consumer Sols., LLC,\nNo. 1:15-cv-01614 (N.D. Ill. Feb. 23, 2015), Dkt. No. 1 (misrepresenting ability to detect melanoma by analyzing pictures of consumers' skin); Compl.,\nIn re Sears Holdings Mgmt. Corp.,\nFTC Docket No. C-4264 (Sept. 9, 2009) (tracking consumers' “online browsing”).\n\n30.\nSee,\ne.g., FTC\nv.\nLendingClub Corp.,\nNo. 18-CV-02454-JSC, 2020 WL 2838827, at *6 (N.D. Cal. June 1, 2020) (inadequate disclosures of loan\norigination fees); see also\nFTC\nv.\nCyberspace.Com LLC,\n453 F.3d 1196, 1200 (9th Cir. 2006) (“A solicitation may be likely to mislead by virtue of the net impression it creates even though the solicitation also contains truthful disclosures.”).\n\n32.\nSee,\ne.g.,\nConsent Order,\nIn re Workado, LLC,\nNo. C-4822 (F.T.C. Aug. 21, 2025) (deceptive claims regarding accuracy or efficacy of its AI-content detection products),\nhttps://www.ftc.gov/system/files/ftc_gov/pdf/ContentatScaleAI-DecisionandOrder.pdf;\nStipulated Order,\nFTC\nv.\nTheFBAMachine Inc.,\nNo. 24-cv-6635-JXN-LDW (D.N.J. July 31, 2025), Dkt. No. 162 (falsely guaranteeing consumers could make money operating online storefronts using AI-powered software); Consent Order,\nIn re DoNotPay, Inc.,\nFTC Docket No. C-4812 (Jan. 14, 2025) (deceptive claims about the ability of its AI chatbot to replace the services of a human lawyer),\nhttps://www.ftc.gov/system/files/ftc_gov/pdf/2323042_donotpay_decision_and_order_0.pdf;\nConsent Order,\nIn re Intellivision Techs. Corp.,\nNo. C-4809 (F.T.C. Jan. 8, 2025) (deceptive claims regarding accuracy or efficacy of its AI-powered facial recognition software),\nhttps://www.ftc.gov/system/files/ftc_gov/pdf/2323023c4809intellivisionfinalorder.pdf;\nStipulated Order,\nFTC\nv.\nEvolv Techs. Holdings, Inc.,\nNo. 1:24-cv-12940-PGL (D. Mass. Nov. 26, 2024), Dkt. No. 2 (false claims about the efficacy of its AI-powered security screening system for detecting weapons).\n\n33.\nSee Compl.,\nFTC\nv.\nAir Ai Techs., Inc.,\nNo. 2:25-cv-03068-SMB (D. Ariz. Aug. 25, 2025), Dkt. No. 1 (deceptive claims regarding efficacy and profitability related to the company's “conversational AI” product).\n\n34.\nIn fact, the Commission has already taken steps to address these unnecessary burdens, for example, by setting aside the previous Commission's order against a generative AI company.\nSee\nOrder Reopening and Setting Aside Order at 5,\nIn re Rytr LLC,\nFTC Docket C-4806 (Dec. 22, 2025),\nhttps://www.ftc.gov/system/files/ftc_gov/pdf/Rytr-Order.pdf\n(“Where actors use AI to violate the law or deceive consumers about the capabilities of their generative AI, they should be held accountable, as the FTC has done and will continue to do. But that is not the case here. Rushing in to impose aggressive law enforcement unsupported by facts or law is improper and is not in the public interest.”).\n\n35.\nSee,\ne.g.,\nOpenAI,\nWhy Language Models Hallucinate\n(Sept. 5, 2025) (“At OpenAI, we're working hard to make AI systems more useful and reliable. Even as language models become more capable, one challenge remains stubbornly hard to fully solve: hallucinations.”),\nhttps://openai.com/index/why-language-models-hallucinate/;\nAnthropic,\nClaude Product Overview, https://claude.com/product/overview\n(last visited May 22, 2026) (“Meet your thinking partner . . . Tackle any big, bold, bewildering challenge with Claude . . . The AI for problem solvers . . . Tackle your toughest work . . . Like an expert in your pocket . . . There's never been a worse time to be a problem, or a better time to be a problem solver . . . What problem are you up against?”); Anthropic,\nHow People Ask Claude for Personal Guidance, https://www.anthropic.com/research/claude-personal-guidancehttps://www.anthropic.com/research/claude-personal-guidance\n(Apr. 30, 2026) (“Helpfulness is one of Claude's most important traits. Speaking with Claude should be akin to a conversation with a brilliant friend, one who will speak frankly to a person about their situation, providing information grounded in evidence.”); X.ai,\nChat With Grok, https://x.ai/grok\n(last accessed May 20, 2026) (“Grok is your truth-seeking AI companion for unfiltered answers with advanced capabilities in reasoning, coding, and visual processing.”).\n\n36.\nSee,\ne.g.,\nAnthropic,\nClaude, https://claude.com\n(last visited May 22, 2026) (“What is Claude and how does it work? Claude is an artificial intelligence, trained by Anthropic using Constitutional AI to be safe, accurate, and secure—the trusted assistant for you to do your best work. . . . What should I use Claude for? If you can dream it, Claude can help you do it. Claude can process large amounts of information, brainstorm ideas, generate text and code, help you understand subjects, coach you through difficult situations, simplify your busywork so you can focus on what matters most, and so much more.”); OpenAI,\nChatGPT Capabilities Overview https://help.openai.com/en/articles/9260256-chatgpt-capabilities-overview\n(last visited Mar. 17, 2026) (touting ChatGPT's ability to solve problems); OpenAI,\nExtracting Insights with ChatGPT Data Analysis, https://help.openai.com/en/articles/9213685-extracting-insights-with-chatgpt-data-analysis\n(last visited Mar. 17, 2026) (claiming ChatGPT can identify missing or outlier data); OpenAI,\nChatGPT for Healthcare, https://help.openai.com/en/articles/20001046-chatgpt-for-healthcare\n(last visited Mar. 17, 2026) (“ChatGPT for Healthcare can pull from millions of peer-reviewed studies, clinical guidelines, and public health sources” to “[g]et clinical answers with citations” that “perform[] better than human baselines across every role measured”).\n\n37.\nSimilarly, an ability to make accurate predictions, correctly solve problems, or make good decisions hinges on AI systems operating without undisclosed ideological biases that contradict explicit and implicit claims of accuracy and neutrality. Though the exact line of what constitutes bias may be difficult to draw, forcing an AI system to prioritize a singular ideological objective over all else is on the wrong side. See Concurring Statement of Comm'r Andrew N. Ferguson,\nIn re IntelliVision Techs. Corp.,\nMatter No. 2323023, at 1 (F.T.C. Dec. 3, 2024),\nhttps://www.ftc.gov/system/files/ftc_gov/pdf/intellivision-ferguson-concurrence.pdf.\n\n40.\nReasonable consumers would not expect, of course, that an AI system would output content that is clearly and unequivocally not protected by the First Amendment, such as child pornography. And nothing in this statement shall be construed to prohibit companies from imposing limits on model use to prevent cybersecurity attacks.\n\n44.\nSee E.O. 14281, Restoring Equality of Opportunity and Meritocracy, at sec. 1, 90 FR 17537 (Apr. 23, 2025) (“Disparate-impact liability all but requires individuals and businesses to consider race and engage in racial balancing to avoid potentially crippling legal liability. . . . On a practical level, disparate-impact liability has hindered businesses from making hiring and other employment decisions based on merit and skill, their needs, or the needs of their customers because of the specter that such a process might lead to disparate outcomes, and thus disparate-impact lawsuits. . . . Disparate-impact liability imperils the effectiveness of civil rights laws by mandating, rather than proscribing, discrimination.”). Indeed, it seems likely a law restricting truthful speech because it might lead another person to commit disparate impact discrimination would not survive First Amendment scrutiny, but that is orthogonal to the FTC Act issues we address here.\n\n46.\nThis conduct is distinct from the problem of incorrect output from AI systems, frequently called “hallucinations,” that stem not from a design decision to prioritize objectives contrary to users' reasonable expectations, but from the technological and resource limitations AI systems necessarily reflect. While a company may, for example, unlawfully deceive consumers if it misrepresents the likelihood of such hallucinations, the Commission does not believe such hallucinations in and of themselves raise issues under section 5 or any other law the Commission enforces.\n\n49.\nSchneidewind\nv.\nANR Pipeline Co.,\n485 U.S. 293, 300 (1988) (“[S]tate law is pre-empted when it actually conflicts with federal law. Such conflict will be found when it is impossible to comply with both state and federal law, or where the state law stands as an obstacle of the full purposes and objectives of Congress.” (internal quotations omitted)).\n\n50.\nFTC\nv.\nDirect Mktg. Concepts, Inc.,\n624 F.3d 1, 12 (1st Cir. 2010) (explaining disclaimers “are not adequate to avoid liability unless they are sufficiently prominent and unambiguous to change the apparent meaning of the claims and to leave an accurate impression”));\nRemovatron Int'l Corp.\nv. FTC, 884 F.2d 1489, 1497 (1st Cir.1989) (“[d]isclaimers or qualifications in any particular ad are not adequate to avoid liability unless they are sufficiently prominent and unambiguous to change the apparent meaning of the claims and to leave an accurate impression.”).\n\n51.\nSee,\ne.g.,\nDeception Policy Statement at *48,\nsupra\nnote 18 (“Depending on the circumstances, accurate information in the text may not remedy a false headline because reasonable consumer may glance only at the headline.”); see also\nFTC\nv.\nCorpay, Inc.,\n164 F.4th 807, 836 (11th Cir. Jan. 6, 2026) (disclaimers or qualifying language may not correct a misleading impression when the disclaimer “is small, ambiguous, or contradicted by the body of the ad”) (citing\nFTC\nv.\nCyberspace.com LLC,\n453 F.3d 1198, 1200-01 (9th Cir. 2006);\nFTC\nv.\nDirect Mktg. Concepts, Inc.,\n624 F.3d 1, 12 (1st Cir. 2010) (explaining disclaimers “are not adequate to avoid liability unless they are sufficiently prominent and unambiguous to change the apparent meaning of the claims and to leave an accurate impression”);\nRemovatron Int'l Corp.\nv.\nFTC,\n884 F.2d 1489, 1497 (1st Cir.1989) (“[d]isclaimers or qualifications in any particular ad are not adequate to avoid liability unless they are sufficiently prominent and unambiguous to change the apparent meaning of the claims and to leave an accurate impression.”).", "url": "https://wpnews.pro/news/policy-statement-concerning-the-suppression-of-accuracy-in-ai-systems", "canonical_source": "https://www.federalregister.gov/documents/2026/07/07/2026-13628/policy-statement-concerning-the-suppression-of-accuracy-in-artificial-intelligence-systems", "published_at": "2026-07-08 22:33:46+00:00", "updated_at": "2026-07-08 22:42:13.232923+00:00", "lang": "en", "topics": ["ai-policy", "ai-ethics", "artificial-intelligence"], "entities": ["Federal Trade Commission", "Federal Register"], "alternates": {"html": "https://wpnews.pro/news/policy-statement-concerning-the-suppression-of-accuracy-in-ai-systems", "markdown": "https://wpnews.pro/news/policy-statement-concerning-the-suppression-of-accuracy-in-ai-systems.md", "text": "https://wpnews.pro/news/policy-statement-concerning-the-suppression-of-accuracy-in-ai-systems.txt", "jsonld": "https://wpnews.pro/news/policy-statement-concerning-the-suppression-of-accuracy-in-ai-systems.jsonld"}}