FTC’s policy statement on “suppression of accuracy” in AI systems takes aim at states’ efforts to regulate AI: What it means for businesses

Executive Summary

What happened: On July 7, 2026, the Federal Trade Commission (FTC) published a proposed policy statement taking aim at states’ efforts to regulate artificial intelligence (AI). The FTC declared that AI companies that steer system outputs “contrary to consumers’ reasonable expectations … including attempted compliance with a State law, such as Colorado’s recently revised Artificial Intelligence Act … may deceive consumers in violation of Section 5 of the FTC Act.”¹ Per the FTC, its policy statement was directed by Executive Order 14365 and carries a short public comment period ending on July 31, 2026.

What it means: The FTC is deploying the same enforcement playbook it has used for two decades in data privacy: Identify reasonable consumer expectations, find undisclosed conduct that contradicts them and regulate it as deception. Companies that quietly modify AI outputs—whether to advance ideological goals, comply with state laws like Colorado’s AI Act² or pursue any other undisclosed objective—face potential Section 5 liability. The FTC also signals a theory of implied federal preemption over state AI laws that require conduct the agency considers deceptive.³ Historically, the FTC has signaled its future enforcement intentions by releasing policy statements and other regulatory guidance materials.

Who it affects: Any company that markets AI systems—from foundation model developers to deployers and enterprise users making AI performance representations to customers.

Practical implications: Companies may wish to review their AI systems for undisclosed output steering, review marketing representations, evaluate whether state-law compliance measures introduce undisclosed accuracy trade-offs, develop a disclosure strategy that meets the FTC’s high “clear and conspicuous” bar and consider filing comments by July 31.

The FTC as America’s De Facto AI and Privacy Enforcer

With no general federal privacy law, the FTC has served as federal privacy and consumer data protection regulator by leveraging its Section 5 authority to police “unfair or deceptive acts or practices.”

Starting in the early 2000s, the FTC brought consent decrees against large educational, pharmaceutical and technology companies for broken privacy promises and inadequate data security. Over the next two decades, it brought hundreds of enforcement actions—against a wide range of companies in many industries —building an extensive body of “common law” establishing that (1) companies’ representations create consumer expectations and (2) undisclosed divergence from those expectations constitutes deception. More recently, the FTC has gone so far as to characterize as data security breaches any uses of consumer data that do not align with companies’ disclosures and consumer consents obtained.⁴

The AI policy statement follows the exact same playbook: AI companies represent that their systems aim to be accurate and helpful, consumers rely on that representation and undisclosed steering violates those expectations. The FTC’s recent AI-specific enforcement actions all relate to the FTC’s concern about misrepresentations of AI capabilities. This policy statement is the framework for an expanded enforcement program.

What the Policy Statement Says

The core theory in plain English: If you steer AI outputs away from what users reasonably expect—and don’t tell them—that’s deception, says the FTC, under Section 5.
The FTC applies its standard three-part deception test: a representation or omission (1) likely to mislead—here, the implicit promise that AI systems aim to produce the best output possible, (2) judged from the perspective of a reasonable consumer, as more than 90% of users accept AI outputs without fact checking, and (3) material to the consumer’s decision to use or pay for the service.

What’s covered: Deliberate design decisions to suppress accuracy—training models to embed ideological distortions, modifying outputs for political agendas, sacrificing accuracy for “equity” or disparate impact avoidance or altering outputs to comply with state laws—all without disclosure.

What’s carved out: Hallucinations (technological limitations, not design choices); balancing accuracy with succinctness, clarity and relevance; and blocking clearly illegal content (cybersecurity assessment and management, cyberattack instructions). Pursuing multiple objectives consistent with user expectations is fine.

Motives are irrelevant: Whether driven by profit, ideology or state law compliance, deception is prohibited by Section 5 regardless.

Who Is Covered

The policy applies to “companies that market artificial intelligence systems,” defined broadly as:

• foundation model developers, and the FTC specifically names a number of companies in this policy statement 
• application builders marketing AI-powered products to consumers or businesses
• deployers and enterprise users who make representations about AI-powered offerings

The trigger is simple; if your marketing—explicitly or implicitly—creates the expectation that your AI system aims to produce the best output for the user, you are within scope.

The Disclosure Safe Harbor

Companies can avoid liability by disclosing that their systems prioritize objectives other than raw accuracy. But the FTC sets a high bar:

• must be “clear and conspicuous”—cannot be buried in terms of service
• must be prominent, persistent and placed where users will actually see it
• must affirmatively dispel the notion the system gives the best possible answer
• the more the disclosure contradicts reasonable expectations, the more prominent it must be
• one-time fine-print disclaimers are likely insufficient
• a “prominent misrepresentation is unlikely to be remedied by a less prominent, subsequent disclosure”

The practical catch: Disclosing that your AI system deprioritizes accuracy to achieve other goals may undermine your product’s value proposition. The FTC effectively forces a choice: Disclose (and potentially harm your market position) or face enforcement risk.

State Law Preemption

The policy statement’s most consequential long-term implication may be its treatment of state AI laws—particularly Colorado’s Artificial Intelligence Act (both original SB 24-205 and revised SB 26-189), which requires companies to explain to the consumer any “adverse decisions” made by the AI system and gives the consumer the right to seek human redress.

The FTC’s position is that although the FTC Act does not expressly preempt state law, state law is impliedly preempted where it conflicts with the federal regulatory scheme. A state law that requires an AI company to suppress accuracy without disclosure—i.e., to deceive consumers—conflicts with Section 5’s core purpose. State-law compliance is not a defense to federal deception claims.

The compliance dilemma: Companies operating in Colorado (and potentially other states with similar AI liability regimes) may face a quandary—comply with state law by modifying outputs, potentially triggering FTC liability, or prioritize accuracy, potentially triggering state liability. The FTC’s answer: Disclose. But as noted above, meeting the FTC’s standard for adequate disclosure is a high bar, and it may undermine the very product a company is selling.

Practical Guidance and Recommended Actions

To prepare for potential heightened FTC enforcement related to AI systems, there are a number of steps companies may want to consider taking. Among these possible steps are the following:

1. Audit for undisclosed output steering. Identify instances in which AI systems steer outputs away from user objectives for reasons beyond technological limitations—such as bias mitigation, equity adjustments or state-law compliance.
2. Review marketing representations. Assess explicit and implicit claims about AI accuracy and capabilities. View all representations through the lens of a possible FTC characterization of “deception” under Section 5.
3. Evaluate state-law compliance measures. If you’ve modified AI outputs to comply with Colorado or other state AI laws, determine whether those modifications are adequately disclosed to users.
4. Develop a disclosure strategy. If your AI systems prioritize objectives other than accuracy, build clear, conspicuous, persistent disclosures that meet the FTC’s proposed standard—and pressure-test whether those disclosures undermine your product.
5. Document design decisions. Maintain records distinguishing technological limitations from deliberate output modifications, with documented business justifications.
6. Engage counsel on the state/federal conflict. Seek guidance on navigating the tension between state AI liability laws and this proposed federal enforcement position.
7. Consider filing comments by July 31. Companies materially affected may want to weigh in—particularly on the disclosure standard, multi-objective AI systems and the preemption theory.

This is a proposed policy statement—not a final rule—and may change based on comments received. But given the FTC’s track record of following through on enforcement positions signaled in policy statements, companies may want to treat this as a strong indicator of the commission’s enforcement priorities.

Comment Period

Deadline	July 31, 2026
Docket	FTC-20260859
File Online	regulations.gov (search docket FTC-2026-0859
Reference	"AI Policy Statement; Matter No. P264200

___________

If you have any questions about this Legal Briefing, please feel free to contact any of the attorneys listed or the Eversheds Sutherland attorney with whom you regularly work.

_{1 91 Fed. Reg. 128 at page 41638.}

_{2 See, https://www.eversheds-sutherland.com/en/united-states/insights/ctrl-alt-legislate-colorado-reboots-its-ai-act}

_{3 The FTC does not note any specific Colorado law provisions that require output steering and instead explains that since there are no “State law safe harbors” within the FTC Act, companies must comply with Section 5 of the FTC Act and its policy statement on deception—“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.” https://www.ftc.gov/legal-library/browse/ftc-policy-statement-deception}

_{4 https://www.ftc.gov/legal-library/browse/cases-proceedings/2023090-goodrx-holdings-inc}