AINEWS
A July proposal from the Federal Trade Commission would let the agency punish AI companies for outputs it calls inaccurate. Digital rights groups say it could override state law.
On July 7, the Federal Trade Commission published a short document that critics say could let the agency override state consumer-protection and civil-rights laws in the name of "AI accuracy." The proposed policy statement treats an AI company's choice to suppress "accuracy" in its own outputs as a potential "deceptive practice" the FTC can punish.
A "policy statement" is not a final rule. It is guidance the FTC uses to telegraph how it will enforce the law, and the agency is collecting public comment before finalizing. The FTC's own press release frames the proposal as a response to a concern that AI companies may be "tempted" to suppress conservative viewpoints or true factual claims. The full FTC policy PDF is the working text the agency will revise based on public comments.
The mechanism is narrower than "AI speech" rhetoric suggests, and broader than the agency is selling. A company that refuses to generate a category of content — for harassment, child-safety, or anti-discrimination reasons — could be argued to be "suppressing accuracy" under the proposal's language. State laws that require such refusals, or that create consumer-protection liability for AI outputs, could then be framed as interfering with the FTC's preferred version of accuracy.
Free Press policy counsel Shilpa Jindia filed a statement on August 3, 2026 calling the proposal part of a "tech-Trump alliance" that "furthered the administration's corruption and enabled its campaign to censor and control our information ecosystem." She argued the proposal would let the FTC coerce companies into "promoting white nationalist narratives." That is a critic's characterization, not a fact the FTC has accepted in the public record; the agency has framed the proposal as a consumer-protection move, not a speech-control one.
The Electronic Frontier Foundation's formal comments push a different point: the language is so broad it could be turned against any AI product that declines to generate content on demand, including outputs that would violate state-level civil-rights, harassment, or consumer-protection rules.
That is the bridge to a concrete case. In March 2026, three young girls sued Elon Musk's xAI, alleging its Grok chatbot was used to create child sexual abuse material. xAI then sued to block Minnesota's nude-deepfake ban, which restricts AI-generated sexual images used to harass or abuse real people. A federal judge denied xAI's request to block the law last week. The Guardian has separately reported on Grok's "undress" image feature and the harassment cases that followed.
The Minnesota law is a state consumer-protection and civil-rights rule. If the FTC finalizes a policy that treats an AI company's refusal to generate content as "suppression of accuracy," the federal guidance could be cited to override state rules like Minnesota's — not by Congress, but by agency interpretation. That is the preemption mechanism Jindia and EFF both flagged. It is not settled law. The proposal is still in the public-comment phase.
President Trump has issued multiple executive orders attempting to shape AI development and discourage state-by-state regulation. The FTC's policy statement would be one tool in that push. Its final form depends on what the agency hears during the comment window — which is the public-record moment digital-rights groups are now using to make the preemptive case.
A reader who has never read an FTC filing can act on this without becoming one. The Federal Register notice, the FTC press release, and the EFF filing all link to the same public comment docket. The deadline, the agency's response to critics, and any revised draft will land in the same record.