On June 1, 2026, Maryland's SB 141 took effect, making the state — by Public Citizen's count — the 30th in the country to criminalize AI deepfakes that spread election misinformation. Signed by Governor Wes Moore as Chapter 444, the law cleared the Senate 44–0 and the House 95–35 on its way through Annapolis (Maryland General Assembly). With midterms approaching, lawmakers wanted a tool in place before the first synthetic robocall or fabricated candidate video hits voters' phones.
What the law does
SB 141 prohibits a person from "knowingly or with reckless disregard creating, using, or disseminating a deepfake to produce materially false information" about an election. It directs the State Administrator of Elections to act on credible reports — and, critically, to "publicly communicate correct information." It authorizes circuit courts to order online platforms to remove offending content. Violations are a misdemeanor carrying penalties up to $5,000 and five years. The statute carves out satire, parody, and bona fide news reporting, with a disclosure requirement for AI-generated content presented as such.
The case for it, stated fairly
The strongest argument for Maryland's law is concrete and recent. Generative tools now let a single actor fabricate a convincing video of a candidate conceding, confessing, or telling supporters the wrong polling date — and push it to millions before any correction can catch up. Defamation suits move in months; an election is decided in hours. A statute that targets knowing or reckless fabrication, exempts satire and journalism, and pairs takedowns with an affirmative duty to publish the truth is a serious attempt to protect the integrity of the franchise, not a blanket speech ban. Ben Yelin of the University of Maryland, who testified in favor, acknowledged in committee testimony that deepfake laws face real First Amendment hurdles — and argued SB 141 was drafted to survive them.
Why California is the warning
That optimism collides with precedent. In 2024 California passed AB 2839, a broader election-deepfake law. Christopher Kohls — a satirist known online as "Mr. Reagan," whose AI-edited Kamala Harris video drew over 100 million views — sued. In October 2024 the court preliminarily enjoined the law, and on August 29, 2025, Senior District Judge John Mendez granted summary judgment and permanently enjoined it.
Mendez's reasoning matters for Maryland. He found AB 2839 was a content-, viewpoint-, and speaker-based restriction on "core political speech" that failed strict scrutiny because it was not narrowly tailored. He pointed out that defamation, privacy torts, and copyright "already provide recourse," and warned that "civil penalties for criticisms on the government… have no place in our system of governance." Tellingly, he sketched a constitutional path: a law confined to "false speech that actually causes voter interference, coercion, or intimidation" — the mechanics of voting, not unflattering depictions of candidates — could survive.
Where Maryland is better — and where it is exposed
Maryland learned part of that lesson. Its disclosure-based satire and news carve-outs are cleaner than California's, and the affirmative correction duty is exactly the kind of counter-speech remedy courts prefer to censorship. The criminal threshold — "knowingly or with reckless disregard" — is higher than AB 2839's, narrowing the law toward genuine bad actors.
But two design choices remain exposed. First, the law still reaches deepfakes that "produce materially false information" generally, not solely false information about when, where, and how to vote — the narrow lane Mendez endorsed. To the extent it sweeps in deceptive-but-political depictions of candidates, it invites the same content-discrimination finding that sank California's statute. Second, court-ordered platform removals are a prior-restraint problem in waiting. Prior restraints carry the heaviest presumption of unconstitutionality in American law, and Section 230 generally shields platforms from being treated as the publisher of user content. A takedown order aimed at the host, rather than liability aimed at the fabricator, is the most legally fragile part of the machinery.
The proportionate path
None of this means Maryland should do nothing. The pro-innovation position is not deregulation — it is precision. The parts of SB 141 most likely to survive judicial review are also the parts that best serve voters: the duty to rapidly publish accurate information, and liability narrowly trained on deliberate fabrications about the voting process itself. The parts most likely to be enjoined — broad "materially false" coverage of candidate depictions and compelled platform takedowns — are also the ones that chill satire, commentary, and the messy political speech the First Amendment exists to protect.
Thirty states now have election-deepfake laws on the books, and a federal court has already shown which version holds up. Maryland's is better drafted than the one that lost. Whether it is narrow enough will likely be settled not in Annapolis, but in a courtroom — and the precedent is not on the side of the broadest reading.