On July 1, 2026, the Federal Trade Commission proposed a policy statement that could reshape how AI vendors think about state-level bias regulations. The statement argues that AI companies steering their systems' outputs for undisclosed ideological reasons violate Section 5 of the FTC Act. That part drew relatively little controversy. What drew significant attention from compliance teams was a different claim buried in the document: that Colorado's AI Act may be impliedly preempted by federal law.
The public comment period closes July 31, 2026. That is four days away. Teams with Colorado operations need to understand what this actually means before the window closes.
TL;DR: The FTC's July 1 proposed policy statement says AI output steering for hidden ideological goals is consumer deception under Section 5. The same document claims Colorado's AI Act (SB 205) is "impliedly preempted" where it pressures vendors to alter accurate outputs to avoid disparate impact liability. The preemption theory is legally untested, and Colorado's law stays in effect until a court says otherwise. Public comment deadline: July 31, 2026.
The statement was issued by FTC Chairman Andrew N. Ferguson, acting pursuant to Executive Order 14365, which President Trump signed on December 11, 2025. That order directed the FTC to clarify how Section 5 applies to AI models, specifically how state laws requiring changes to AI outputs can conflict with federal consumer protection law.
What the statement actually says
The core of the proposed policy statement is that AI systems are consumer products, and consumers rely on them for accurate, unmanipulated outputs. A company that secretly tunes its AI to favor certain viewpoints, suppress certain answers, or weight results in undisclosed ways is, in the FTC's framing, deceiving consumers about the nature of the product they are using. That is the Section 5 claim.
The FTC is not arguing that all AI output curation is deceptive. Companies are allowed to build AI systems that do specific things. A customer service chatbot that only answers questions about your products is fine. The concern is about hidden manipulation that consumers would object to if they knew it was happening -- systems that claim to be neutral or accurate while being quietly adjusted to produce preferred outcomes.
This part of the statement is relatively uncontroversial. Most AI governance professionals would agree that undisclosed output steering raises legitimate consumer protection questions. The preemption argument is where things get more contested.
The Colorado preemption claim
Colorado's Artificial Intelligence Act (SB 205, often called SB 189 in early drafts) requires developers and deployers of high-risk AI systems to take "reasonable care" to protect consumers from known or reasonably foreseeable algorithmic discrimination. If a high-risk AI system produces outcomes with disparate impact across protected classes, the developer and deployer can face liability under the statute.
The FTC statement argues that this creates a conflict with federal law. The reasoning goes like this: to avoid disparate impact liability under Colorado law, AI vendors might alter their systems' outputs -- adjusting who gets flagged, who gets a higher score, who receives a recommendation -- in ways that make those outputs less accurate in a technical sense. That output adjustment, in the FTC's view, is itself a form of output suppression that could violate Section 5. Therefore Colorado's law, by pressuring vendors to suppress accuracy, conflicts with the federal consumer protection framework and is impliedly preempted.
The FTC's statement uses the phrase "impliedly preempted to the extent it conflicts with the federal regulatory scheme established by Section 5 of the FTC Act."
Why this theory is legally contested
Implied preemption is a well-established legal doctrine, but the bar for it is high. Courts require a showing that it is genuinely impossible to comply with both the state and federal requirement simultaneously, or that the state law stands as an obstacle to the accomplishment of the federal purpose.
Several legal commentators have pointed out that the FTC's preemption argument here has not been tested in any court. The FTC Act does not expressly preempt state AI laws. The statement acknowledges this and rests its argument entirely on conflict preemption theory. To prevail, a court would need to find that complying with Colorado's disparate impact standard and complying with the FTC's deception framework are mutually exclusive -- that there is no way to reduce discriminatory AI outputs without also deceiving consumers about accuracy.
That is not an obvious conclusion. Many AI practitioners would argue that reducing disparate impact and maintaining output accuracy are compatible goals, not opposing ones. A more accurate model that does not systematically undercount qualified candidates from a particular demographic group can be both more accurate and less biased simultaneously.
The FTC's preemption argument may be more of a policy signal than a legally settled position. It signals the current administration's view of the federal-state AI regulatory relationship. Whether courts accept it is a separate question.
Does this preemption claim apply to other state AI laws?
The statement singles out Colorado specifically. But the same preemption logic, if accepted by courts, could apply to other state AI laws that impose disparate impact analysis requirements on AI vendors.
Texas TRAIGA (the Texas Responsible AI Governance Act) and Illinois AI legislation both contain provisions addressing algorithmic discrimination and protected-class outcomes. If the FTC's preemption theory succeeds in Colorado litigation, challengers would likely use the same argument against those statutes.
This is why the July 31 comment deadline matters even for teams that do not operate in Colorado. The comments submitted to the FTC record will shape the legal arguments used if this preemption theory is litigated. Industry groups, civil rights organizations, and state governments are all filing comments. Enterprise teams that have invested in multi-state AI compliance programs have a direct interest in how the FTC responds to those comments.
What teams should do before July 31
First, do not stop complying with Colorado's AI Act. The FTC statement is proposed guidance from the executive branch. It is not a court ruling, not a statute, and not a regulation that has completed notice-and-comment rulemaking under the Administrative Procedure Act. Colorado's law remains legally in effect for covered systems. Stopping compliance on the basis of an untested preemption theory would create significant legal exposure.
Second, if you are a covered developer or deployer under Colorado SB 205 and you have built a compliance program, document that program carefully now. If this preemption argument is litigated, having a documented record of your reasonable care efforts under the Colorado framework will be relevant regardless of which way courts rule.
Third, consider filing a comment. The FTC docket is open until July 31. Public comments on proposed policy statements influence how the FTC frames its enforcement priorities, even if the statement itself is not binding regulation. If your compliance program was designed to reduce bias without sacrificing accuracy -- which is what most responsible AI programs aim for -- that practical experience is relevant to the FTC's analysis.
Fourth, watch the response from Colorado's Attorney General and the civil rights organizations that supported SB 205. They are almost certain to file comments arguing against the preemption position, and their arguments will frame the litigation posture if challengers try to use the FTC statement to attack state enforcement.
The broader context
The FTC's statement is part of a pattern in the current administration of using federal agency guidance to challenge state-level AI regulation. The One Big Beautiful Bill Act, passed in May 2026, included a ten-year moratorium on state AI regulation in certain categories. The FTC's preemption argument takes a different legal route to a similar destination: using existing federal consumer protection authority to displace state-level approaches to algorithmic accountability.
Both strategies reflect a view that a fragmented state-by-state AI regulatory framework creates compliance burdens for AI vendors and that federal uniformity should prevail. Whether that view prevails in court, in Congress, or in public discourse is still being determined. The July 31 comment period is one of the venues where that determination happens.
Related Reading
- Colorado AI Act Compliance Deadline: What Teams Must Do Before It Takes Effect
- FTC AI Enforcement Actions 2026: What the Pattern Means for Small Teams
- FTC AI Accuracy Policy: 5 Questions to Ask Your AI Vendor Now
- Texas TRAIGA Compliance Checklist 2026
- Illinois AI Employment Disclosure Law 2026: What HR Teams Must Do
- US State AI Law Tracker: Every Law with Deadlines and Obligations
Sources: FTC press release, July 1, 2026, Federal Register 2026-13628, Inside Privacy analysis, Consumer Finance Monitor analysis.
