The FTC’s 30-day comment window on its proposed AI accuracy policy statement closes July 31, and the docket so far tells a quiet story: 40 comments filed on Regulations.gov docket FTC-2026-0859. That’s a thin record for a statement that, once finalized, would give the Commission a Section 5 deception theory purpose-built to challenge state AI regulation.

The proposal was published July 1 under File No. P264200 and cleared the Commission on a 2-0 Republican vote. Federal Register notice 2026-13628 ran July 7. Its origin isn’t procedural drift, it’s Executive Order 14365, signed by President Trump on December 11, 2025, which directed the FTC to address state laws requiring alteration of “truthful outputs of AI models.”

The legal architecture is compact. The FTC argues AI companies have made explicit and implicit representations that their systems produce the most accurate output available within technological limits. Consumers, per the statement, accept AI outputs without independent fact-checking more than 90% of the time. Steering outputs toward undisclosed objectives, whether “of its own volition or in response to a state law requirement,” may therefore be deceptive under Section 5’s three-part test. Hallucinations from technical limits are carved out. Intentional configuration isn’t.

Colorado’s Artificial Intelligence Act is named. The statement concludes its disparate-impact liability pressures developers to alter outputs, and that such a regime is “impliedly preempted to the extent it conflicts with a federal regulatory scheme.” Companies can dodge Section 5 exposure through clear and conspicuous disclosures that a system prioritizes certain objectives over user expectations.

The maneuver rhymes with the OCC’s 2020s use of preemption doctrine against state consumer-finance rules: an agency policy statement doing the constitutional work Congress hasn’t. The disclosure carve-out is the tell. It’s not a rule against steering. It’s a rule against steering quietly.

Sources

Sources