The Statement That Named the State by Name

On July 1, 2026, the Federal Trade Commission issued a proposed policy statement on the suppression of accuracy in artificial intelligence systems, authorised by a 2-0 Commission vote. The statement does two things at once. It asserts that AI companies which distort model outputs to achieve undisclosed objectives may be engaging in deceptive acts under Section 5 of the FTC Act. And it names a specific state law, Colorado's Artificial Intelligence Act, as impliedly preempted to the extent it conflicts with that federal regulatory scheme. The public comment window closes on July 31.

The statement executes Section 7 of Executive Order 14365, Ensuring a National Policy Framework for Artificial Intelligence, signed on December 11, 2025. That section directs the Chairman of the FTC to issue, within ninety days, a policy statement on the application of the FTC Act's prohibition on unfair or deceptive acts to AI models, and to explain the circumstances under which state laws requiring alterations to the truthful outputs of AI models are preempted. The statement issued on July 1 is that instrument.

What is new about the instrument

Two things.

First, the statement moves the boundary of Section 5 into a domain that has not previously been treated as a federal deceptive-practices question. The FTC's proposed policy statement frames the substantive test as consumer expectation: "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." The statement then describes a specific category of conduct that satisfies deception: an AI company steering outputs "contrary to consumers' reasonable expectations for various reasons, including attempted compliance with a state law." The mechanism of deception, in other words, includes attempts to comply with the very state accountability regimes the Executive Order challenges.

Second, the statement names a specific state statute. The proposed policy statement identifies Colorado's Artificial Intelligence Act as an example of a law that "appears to coerce companies into altering the output of their AI models to comply with and advance the state's ideological objectives" and asserts that it is "impliedly preempted to the extent it conflicts with a federal regulatory scheme." Implied preemption in an agency policy statement is a discretionary framing, not a court finding. It signals the enforcement posture the FTC intends to adopt if a challenge arrives. The Executive Order had already directed the Department of Justice to establish an AI Litigation Task Force whose sole responsibility is to challenge state AI laws inconsistent with the Order's policy. The FTC statement supplies the deceptive-practices framing that Task Force filings may draw upon.

Where the governance object is moving

The policy statement is the fifth instrument in a sequence of federal actions since December 2025 that reallocate substrate authority. EO 14365 directed the Attorney General to establish the litigation task force, directed the Secretary of Commerce to evaluate state AI laws for onerous provisions, directed the Secretary to condition BEAD non-deployment funds on state AI-law posture, directed the FCC to consider a federal reporting and disclosure standard that preempts conflicting state laws, and directed the FTC to issue the policy statement now under review.

Read together, the five directives describe a governance object that is being pulled out of the state civil-rights domain and relocated inside the federal deceptive-practices domain. In the state-level frame, algorithmic accountability laws such as Colorado's require impact assessments, disclosures, and specified duties of care when an automated decision system is used in a consequential domain. The governance object is the consequential decision, and the governance function is anti-discrimination. In the federal deceptive-practices frame the statement is now proposing, the governance object is the output the consumer sees, and the governance function is truth in the marketplace. These are not the same object and they are not the same function. One frame considers whether an automated decision produced an unlawfully discriminatory outcome. The other considers whether an output was steered away from what a consumer would reasonably expect the model to produce.

The proposed policy statement stipulates the swap explicitly. It states that Section 5 applies "even when a company engages in a deceptive act or practice in order to comply with a state law." The federal deceptive-practices regime is being framed as the operative regime, and the state accountability regime is being framed as the source of the deception rather than as a co-equal safeguard. The implied-preemption doctrine is what carries the swap.

What the deployment tempo looks like now

The instrument itself is a proposed policy statement, not a rule and not a court decision. It goes into a thirty-day public comment window running through July 31, 2026. Commission votes on policy statements do not by themselves generate binding federal regulation, and implied-preemption assertions in an agency policy statement do not by themselves invalidate a state law. What the statement does is coordinate three parallel motions. It signals FTC enforcement direction against companies attempting to comply with the identified state regimes. It supplies the deceptive-practices framing for the AI Litigation Task Force established by the same Executive Order. And it runs on the same ninety-day clock as the FCC preemption-standard proceeding and the Commerce Department state-law evaluation and the BEAD conditionality notice.

Colorado's Artificial Intelligence Act, on the state side, took effect on schedule as a first-of-its-kind algorithmic accountability regime. Fifteen or more other states have moved comparable bills through part or all of the legislative process. The instrument tempo on the state side has been years-per-law. The instrument tempo on the federal side, since the Executive Order was signed on December 11, 2025, has been weeks-per-instrument.

What remains on the table

• What is the sequence in which the Colorado AI Act, and the roughly fifteen comparable state statutes, get tested against the implied-preemption framing announced in the policy statement, and by whom?
• If Section 5 deception is the operative federal governance function, what happens to the algorithmic-accountability governance function that Colorado's statute was written to instrument, since the two functions do not measure the same object?
• Does a policy statement issued under a proposed comment window supply enough of a "federal regulatory scheme" for implied preemption to reach a Colorado-style statute, or does the doctrine require the FCC reporting standard and a further FTC rulemaking to actually be in effect?
• How does the deceptive-practices frame interact with the fact that the model developer, the deployer, and the consumer-facing product are commonly three different parties, only one of which can be liable under Section 5 for steering an output?

The governance artifact was retained. The governance function was not. The policy instruments and the deployment tempo are not aligned.