Maryland is now the 30th state with a law targeting election deepfakes, according to Public Citizen's tracker of state legislation. Gov. Wes Moore signed SB 141 on May 12, 2026, and it took effect June 1, 2026 — five months before the November midterms. The law directs the State Administrator of Elections to act on credible reports of AI-generated election misinformation: publish corrective information, seek court orders forcing platforms to remove offending content, and subpoena information about who distributed it. It also makes it a misdemeanor to knowingly or recklessly create or disseminate a deepfake producing materially false election information, punishable by up to $5,000 and five years in prison.
The Case for the Law
Supporters have a genuine problem to point to. State Sen. Katie Hester, the bill's sponsor, cited scenarios like an AI-generated image of a polling place on fire or a fabricated video of immigration agents stationed at a voting location — content designed to suppress turnout in the narrow window before polls close, when there's no time for litigation or platform review to catch up. Courts move slowly; a viral fake can do its damage in hours. Even the federal judge who later struck down California's comparable law agreed that states have a compelling interest in protecting election integrity from synthetic media. That much isn't in dispute.
The Same Design That Just Failed in Court
The problem is structural, not rhetorical. In Kohls v. Bonta, Senior U.S. District Judge John Mendez permanently enjoined California's AB 2839 in August 2025, after first blocking it on a preliminary basis in October 2024, finding it imposed content, viewpoint, and speaker-based restrictions on core political speech and failed strict scrutiny because it wasn't narrowly tailored. The court held that vague standards for what counts as "materially deceptive" chill lawful expression, and that counter-speech — not government-adjudicated takedowns — is the constitutionally preferred remedy for political falsehoods (case docket: govinfo.gov, USCOURTS-caed-2_24-cv-02527).
Maryland's SB 141 shares the same architecture the court objected to: a subjective "materially false information" standard, criminal penalties that attach before any judicial review, and a satire/parody exemption that requires the state to first evaluate a speaker's intent before deciding whether they're exempt. Reason Foundation's testimony to the Maryland legislature made this point directly, warning that the bill "chills political speech" by empowering the government to determine, after the fact, whether a piece of content was legitimate commentary or a prosecutable deepfake — precisely the kind of speaker-by-speaker judgment call that doomed California's law (reason.org testimony). A citizen weighing whether to post a satirical AI image of a candidate, unsure how a prosecutor might later characterize it, has every incentive to simply not post — the textbook definition of a chilling effect, and one that operates even if no one is ever actually charged.
A Narrower Path Was Available — And Ignored
What makes Maryland's choice frustrating is that a First Amendment-compatible alternative already exists next door in the statute books. Utah's HB 329 requires disclosure when AI-generated content appears in paid political advertising — a narrower target that reaches candidates, campaigns, and PACs already subject to campaign-finance disclosure regimes, without criminalizing an ordinary voter's meme or a late-night comedian's parody clip. Reason Foundation's testimony urged Maryland to adopt this model instead; lawmakers declined. The Kohls v. Bonta court itself sketched out where a constitutional line could be drawn: statutes limited to "factual statements that are demonstrably false — like the time, date, place, or manner of voting" survive scrutiny because they regulate objectively verifiable claims, not contested political speech (electionlawblog.org). A law that says "don't lie about when polls close" is categorically different from one that lets an election administrator decide whether an AI-generated ad about a candidate's record is "materially false."
What Comes Next
Maryland's law will likely draw a First Amendment challenge before November, following the pattern set in California, Texas, and Minnesota, where similar statutes have already been enjoined or narrowed by courts. That's not a reason to do nothing about AI-generated election disinformation — the underlying harm is real, and disclosure-based rules for paid political advertising, plus rapid-response fact-checking from election offices, address the acute threat without inviting the state to referee political speech. But broad criminal bans on "materially false" content, adjudicated after the fact by the same officials such content might target, keep losing in federal court for the same reason every time. Thirty states have now legislated on this problem; the ones that modeled their laws on Utah's disclosure approach are far less likely to be next in line for an injunction than the ones that modeled theirs on California's.