Shontel Brown introduces bill requiring AI labels on certain election content

Rep. Shontel Brown introduced the Securing Elections from AI Deception Act on October 21, 2026, proposing national disclosure rules for specified AI-generated election content. The bill would also prohibit developers and deployers from intentionally using AI to deprive or defraud people of voting rights, pairing labels with a separate ban on harmful conduct.

The proposal has 30 House cosponsors, according to Brown’s office. It has endorsements from the NAACP, the Lawyers’ Committee for Civil Rights Under Law and NETWORK Lobby for Catholic Social Justice. Brown first introduced the legislation in 2024; this is a renewed legislative effort, not an enacted requirement.

A sentence or a symbol

The disclosure requirement would cover federal, state and local election content. Brown’s office describes covered material as AI-generated images, text, video, audio or other media containing simulated, altered or distorted depictions of candidates. It would also cover information about the time, place, manner or requirements for voting or registering to vote.

That scope reaches beyond fabricated candidate footage: voting instructions are a separate category of covered information. For material subject to the rule, the proposal offers two ways to disclose AI use:

• A written statement: “AI Disclaimer: This output was generated by artificial intelligence”.
• A symbol consisting of the letters “AI” inside a circle, similar in size and design to the copyright symbol.

The Federal Trade Commission would regulate and enforce the disclaimer requirement through its authority over unfair or deceptive practices. Its role would extend beyond enforcement: the agency would have to educate the public about the disclaimer’s meaning and use, then report to Congress on its adoption and effectiveness.

Intentional interference would face separate penalties

The bill’s voting-rights provisions target intentional conduct by AI developers or deployers. Brown’s summary lists deceptive information about voting times, locations and methods; eligibility to vote or register; ballot counting; election adjudication; endorsements; and voting procedures or requirements. These provisions address how AI is used to interfere with rights, rather than simply whether content carries a label.

The prohibitions also encompass deception, threats, intimidation, fraud or coercion intended to prevent, interfere with or deter protected activities. Those activities include registering to vote, voting, supporting or advocating for a candidate, and serving as an election worker. Attempts to interfere would also be covered, with violations subject to a fine or imprisonment.

The NAACP’s endorsement emphasizes that civil-rights concern. Cedric C. Haynes, its senior vice president of policy and legislative affairs, said Black communities have long faced voter deception and intimidation. He warned that new technology can make those tactics more sophisticated and harder to detect.

69%Identified AI ads without disclosure

Wesleyan Media Project’s September 4, 2026 analysis found no AI-use disclaimer on 69% of the 164 political ads it identified as generated or enhanced with AI.

The disclosure gap is difficult to measure

Brown’s announcement cites that Wesleyan Media Project disclosure finding. The underlying analysis was published in September, so it is an earlier snapshot rather than a new October measurement. Its category also includes AI-enhanced advertising, not just ads made entirely by AI—a distinction relevant to understanding what the researchers counted.

Researchers located ads through news coverage, AI disclaimers and their own monitoring for generated visuals and fabricated depictions of people. Their methodology also incorporated ads reported by AdImpact as containing AI material, along with staff-flagged examples confirmed by the team. The tracking covered broadcast television, connected television and digital advertising.

Wesleyan describes its tally as an undercount because AI use can be hard to identify definitively. The finding therefore measures disclosure among ads the team could identify, rather than establishing how often every AI-assisted political advertisement carries a notice.