KENTUCKY’S NEW “SYNTHETIC MEDIA” ELECTION LAW: A DISCLOSURE FIX OR A FIRST AMENDMENT FIGHT?
Author: Ken Smith, Senior Editor
On October 27, 2025, Kentucky Educational Television (“KET”) reported on a video shared by the chair of the Republican Party in Hardin County, Kentucky, that depicted former President Barack Obama and his wife, Michelle Obama, as primates.[i] The story sparked national coverage, and eventually, the Hardin County chairperson took down their post and offered a public apology.[ii] This form of media, in the electioneering context, is precisely the type of messaging that Kentucky’s Senate Bill 4 (“SB4”) aims to regulate.[iii]
AI video generation has drastically improved in recent years.[iv] Released in September 2025, an OpenAI program called Sora 2 has been praised for its realism and quality in its video generation.[v] Because of programs like this that can produce videos and images with convincing realism, Kentucky sought to regulate their use in electioneering communications.[vi]
What Kentucky Did in Plain English
Following the passage of SB4, any candidate for any public office in Kentucky who experiences an AI-manipulated media communication that alters their appearance or speech can seek an injunction to prohibit the continued distribution of the broadcast or force the sponsor to prominently disclose it as AI-manipulated media.[vii] The Kentucky law offers several guardrails for regulating AI-manipulated media through its definition of terms.[viii] The regulation governs all forms of broadcast media, including cable, social media, telephone messages, and radio.[ix] Kentucky limited the bill only to media broadcast within 45 days before an election day.[x] This narrow window reflects the legislature's intent to protect candidates from last-minute fabrications before ballots are cast.[xi]
Analysis & Potential Constitutional Issues
The choice to require disclosure rather than implement an outright ban follows the lead of many pre-existing state laws throughout the country that have taken the same approach to regulating AI-manipulated electioneering communications.[xii] Compelling disclosure is a meaningfully lighter burden on First Amendment rights than an outright prohibition.[xiii] Despite this more measured approach, SB4 remains vulnerable to a challenge on overbreadth grounds.[xiv]
A law is unconstitutionally overbroad when it prohibits a substantial amount of constitutionally protected speech beyond the unprotected speech it was designed to target.[xv] This concern carries substantial weight in the political speech context because an overly broad law can chill participation in the discourse on which democratic governance depends.[xvi] SB4, despite its narrow purpose, presents three areas in which its reach appears to extend well beyond the AI-generated deepfakes it was designed to address.[xvii]
First, the statute's definition of "synthetic media" appears to be limited to the manipulation of existing audio or visual recordings, which creates a meaningful gap in its coverage of the very technology it seeks to regulate.[xviii] AI programs like Sora 2 can generate entirely original video and audio content from scratch, without any underlying recordings to manipulate or alter.[xix] For overbreadth purposes, however, if courts interpret the definition broadly enough to encompass original content, the statute may simultaneously sweep in constitutionally protected political satire and commentary that has never relied on a real recording but clearly engages with public figures and public affairs.[xx]
The statute's definition of synthetic media also does not specify how much manipulation of an existing recording is required to trigger the disclosure obligation.[xxi] Routine editing choices, including cuts between shots, audio leveling, color correction, and the splicing together of multiple clips, are standard practice in the production of any broadcasted political communication.[xxii] Under a broad reading of the statute, these ordinary production decisions could technically constitute manipulation of audio or visual content, bringing virtually every political advertisement within SB4's reach, regardless of whether any AI technology was involved.[xxiii]
Finally, while SB4 requires that disclosures appear in a "clear and conspicuous" manner so that a viewer would be likely to notice them, it offers no further definition of what those standards entail.[xxiv] Both the FEC and the FCC, in their proposed rulemaking on AI-generated content in political advertising, have found it necessary to prescribe highly specific technical requirements (type size, color contrast, and minimum display duration) to give the phrase a workable meaning in the context of political broadcasting.[xxv] Kentucky's SB4 uses the same phrase but provides none of that specificity, meaning that courts evaluating SB4 compliance will be forced to do so without the regulatory scaffolding that gives the standard coherent content at the federal level.[xxvi] Without that guidance, similarly situated broadcasters may face different outcomes depending on the judge assigned to their case, and that inconsistency in enforcement is a consequence of overbreadth.[xxvii] When the scope of a law's disclosure requirement cannot be reliably predicted in advance, speakers are more likely to self-censor rather than risk noncompliance, and that chilling effect is itself a constitutional harm that courts have long recognized as intolerable.[xxviii]
Together, these features of SB4 suggest that while its core purpose of deterring AI-manipulated smear campaigns in the final days before an election is constitutionally legitimate, its current drafting extends well beyond what that purpose requires.[xxix] A successful overbreadth challenge would not necessarily invalidate the statute in its entirety, as courts may instead apply a narrowing construction that limits SB4's reach to the AI-generated deepfakes it was clearly designed to address.[xxx] Until that is established, however, SB4's language will leave broadcasters, campaigns, and courts without a consistent framework for compliance.[xxxi]
Conclusion
Kentucky SB4 modernizes the Commonwealth’s election laws and will deter the use of AI in election messaging in the latest stages of political campaigns.[xxxii] However, due to the overbroad language and the heavy reliance SB4 will place on the subjective eyes of the controlling judges, affected parties throughout Kentucky will likely see inconsistent decisions. The remaining question is whether any inconsistencies will lead to an overbreadth challenge, or whether the powers that be can hone the few terms in the law that leave a little too much wiggle room.
[i] Hardin Co. GOP Official’s Post Sparks Backlash, Kentucky Edition (PBS Oct. 27, 2025), https://www.pbs.org/video/hardin-co-gop-officials-post-sparks-backlash-5eeifr/.
[ii] Id.
[iii] Ky. Rev. Stat. Ann. § 117.322 (LexisNexis 2025).
[iv] Jamie Vickers, Marc Brodherson, Alec Wrubel & Cléophée Bernard, What AI Could Mean for Film and TV Production and the Industry’s Future, McKinsey & Co. (Jan. 23, 2026), https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/what-ai-could-mean-for-film-and-tv-production-and-the-industrys-future.
[v] Fabian Mosele, AI Timeline: A History of Image and Video Generative Models, Fabian Mosele, https://www.fabianmosele.com/ai-timeline (last visited Feb. 22, 2026).
[vi] Id. See also, Ky. Rev. Stat. Ann. § 117.322 (LexisNexis 2025).
[vii] Id. at § 117.322(1)(a).
[viii] Ky. Rev. Stat. Ann. § 117.001(8)(a) (LexisNexis 2025).
[ix] Id.
[x] Id.
[xi] Id.
[xii] See Political Media, Wash. Univ. in St. Louis Sch. of L., AI Policy & Regulation Resources (noting that the predominant state approach for AI-generated content in political advertising requires disclosure rather than outright prohibition), https://law.washu.edu/ai-policy-and-regulation-resources/political-advertising/.
[xiii] See id.
[xiv] See Broadrick v. Oklahoma, 413 U.S. 601, 612 (1973) (holding that a law is unconstitutionally overbroad when it prohibits a substantial amount of protected speech beyond its legitimate target).
[xv] Id.
[xvi] Citizens United v. Federal Election Commission, 558 U.S. 310, 340 (2010) (reaffirming that political speech occupies the highest rung of First Amendment protection and that laws burdening such speech warrant heightened constitutional scrutiny).
[xvii] Ky. Rev. Stat. Ann. § 117.001(20)(a) (LexisNexis 2025).
[xviii] See id.
[xix] Artlist, Sora 2 Ultra-Realistic Video Generator, Artlist AI, https://artlist.io/ai/models/sora-2 (last visited Feb. 22, 2026).
[xx] See, e.g., Broadrick, 413 U.S. 601, 612 (1973); Citizens United, 558 U.S. 310, 340 (2010).
[xxi] Ky. Rev. Stat. Ann. § 117.001(20)(a) (LexisNexis 2025).
[xxii] Ky. Rev. Stat. Ann. § 117.322 (LexisNexis 2025).
[xxiii] Ky. Rev. Stat. Ann. § 117.001(20)(a) (LexisNexis 2025).
[xxiv] Ky. Rev. Stat. Ann. § 117.322 (LexisNexis 2025).
[xxv] 11 C.F.R. § 110.11 (2024) (requiring that disclaimers in political advertising be presented in a clear and conspicuous manner and specifying that a disclaimer is not clear and conspicuous if it is difficult to read or hear, or if its placement is easily overlooked); Disclosure and Transparency of Artificial Intelligence-Generated Content in Political Advertisements, 89 Fed. Reg. 63,381 (Aug. 5, 2024) (to be codified at 47 C.F.R. pts. 25, 73, and 76) (proposing that visual disclosures of AI-generated content in political advertisements appear in a clear and conspicuous manner sufficient to ensure viewer awareness).
[xxvi] Ky. Rev. Stat. Ann. § 117.322 (LexisNexis 2025).
[xxvii] See Broadrick, 413 U.S. 601, 612 (1973) (establishing that overbroad laws risk chilling constitutionally protected expression through unpredictable and inconsistent enforcement).
[xxviii] See Citizens United, 558 U.S. 310, 340 (2010) (reaffirming that political speech occupies the highest rung of First Amendment protection and that laws burdening such speech warrant heightened constitutional scrutiny); see also Buckley v. Valeo, 424 U.S. 1, 64 (1976) (recognizing that disclosure requirements in the political advertising context serve legitimate governmental interests but must be carefully restrained to avoid unconstitutional burdens on protected speech).
[xxix] See Ky. Rev. Stat. Ann. § 117.322 (LexisNexis 2025); Broadrick, 413 U.S. 601, 612 (1973).
[xxx] United States v. Hansen, 599 U.S. 762, 770 (2023) (holding that a statute should not be deemed facially invalid where it is readily subject to a narrowing construction by the courts, provided its deterrent effect on protected expression is real and substantial).
[xxxi] See Ky. Rev. Stat. Ann. § 117.322 (LexisNexis 2025); see also,Hansen, 599 U.S. 762, 770 (2023).
[xxxii] Ky. Rev. Stat. Ann. § 117.322 (LexisNexis 2025).