Skip to content

State Deepfake-in-Elections Laws: a Dated Snapshot

9 min read · updated August 11, 2026

Most pages on this subject open with a number: how many states have enacted an election-deepfake law. That number is the least reliable thing on the page, because the statutes differ enough that any count depends on what you decided to count. This page lists entries that can each be tied to a named instrument, explains the two shapes those instruments take, and does not assert a total.

Nothing here is legal advice, and this list is explicitly not exhaustive. It is a dated snapshot of entries the author could tie to a primary source at the time of writing. State legislatures amend these provisions every session and courts have enjoined some of them. Verify against the state’s own legislature site before relying on any row.

The two drafting patterns

Nearly every enacted statute in this area is one of two designs, and knowing which one you are looking at tells you most of what you need.

Disclosure

The larger group requires a conspicuous disclosure on synthetic media in election communications — a label saying the content was generated or manipulated — and makes the absence of the label the violation. These statutes typically specify placement and legibility rules (a minimum text size, an audio statement at the beginning of an audio-only communication), attach to a defined window before an election, and are enforced by civil penalty or by an action a candidate may bring.

Disclosure statutes are the constitutionally safer design. Compelled factual disclosure in the election context is reviewed under a more forgiving standard than a prohibition on the content itself, and the state can characterise the burden as informational rather than suppressive.

Prohibition

The smaller group prohibits the dissemination outright, usually conditioned on intent to injure a candidate or to influence the result, a knowledge or recklessness standard, absence of consent, and a window before the election. Minnesota is the clearest example; see the elements of section 609.771. These are the statutes that draw challenges.

A number of states have both: a general disclosure obligation, plus a narrower prohibition for the highest-intent conduct. When someone summarises such a state as “bans election deepfakes”, they have collapsed two provisions with different tests and different remedies into one clause.

Entries that can be sourced

Each of the following is tied to a bill or codified section that can be read at the state’s own site. This is a subset, chosen for traceability rather than for coverage.

StateDescription
TexasSenate Bill 751 (2019) added section 255.004 to the Election Code, creating a criminal offence for publishing a deceptive video within 30 days of an election with intent to injure a candidate or influence the result. Among the earliest in the country. Texas Statutes, Election Code chapter 255.
MinnesotaMinnesota Statutes section 609.771 (2023, amended 2024). Prohibition pattern, 90-day window, criminal penalties plus forfeiture of office on conviction. Subject to a pending federal constitutional challenge. Minnesota Office of the Revisor of Statutes.
WashingtonSenate Bill 5152 (2023) created a synthetic-media-in-campaigns regime built on disclosure, with a civil cause of action for a depicted candidate. Washington State Legislature bill summary. Washington recodified parts of its election law after enactment, so cite the current RCW chapter rather than the bill.
Michigan2023 legislation added a disclosure requirement for materially deceptive media in political advertising, with a criminal offence for the intentional-deception tier. Michigan Legislature.
FloridaHouse Bill 919 (2024) requires a disclaimer on political advertisements containing generative-AI-created depictions of a real person doing something that did not occur, with a criminal penalty for omission. Florida Senate bill page.
Wisconsin2023 Wisconsin Act 123 requires a disclaimer on audio or video communications containing synthetic media in political advertising. Wisconsin Legislative Documents.
CaliforniaAssembly Bills 2839 and 2655 (2024) went further than most, reaching platforms as well as distributors. Both drew immediate federal challenges and neither should be described as simply “in force”; check the current litigation status before citing them. California Legislative Information, AB 2839.

Why published counts disagree

If you compare two trackers you will usually find they differ by several states. The differences are not errors; they are classification choices, and there are four of them.

  • Elections-only or not. Some counts include statutes aimed at non-consensual intimate synthetic imagery or at digital replicas of performers, which are adjacent but regulate different conduct. See California’s digital replica law for one of those.
  • Enacted, effective or enjoined. A bill signed in March with a January effective date is enacted and not yet operative. A statute partially enjoined is on the books and partly unenforceable. Counts rarely say which state they are counting.
  • Synthetic-media-specific or technology-neutral.Several states already had statutes prohibiting materially deceptive campaign communications and simply amended them. Whether that counts as a new deepfake law is a judgement call.
  • Resolutions and study bills. Some trackers include non-binding instruments. Those are not law and should never appear in a compliance count.

This is why the page does not give a total. A number that changes depending on four undisclosed decisions is not a fact; the list of instruments is.

The constitutional pressure

These statutes sit on contested ground, and the position is genuinely unresolved rather than merely uncertain. Political speech is at the core of First Amendment protection, and a content-based restriction on it faces strict scrutiny; the states’ answer is that a knowing, non-consensual, intentionally deceptive depiction of conduct that never occurred is closer to defamation and fraud than to protected commentary. Several federal courts have engaged with that argument and the results have not been uniform, with prohibition-pattern statutes and platform-directed obligations drawing the most scepticism, including on federal preemption grounds unrelated to the First Amendment.

What would settle it is an appellate decision squarely on a disclosure statute and another on a prohibition statute; absent that, the responsible statement is that the disclosure pattern is more likely to survive and that no one should describe either pattern as settled law.

How to check the current position

  1. Start from the state legislature’s own site and find the codified section, not the bill. Bills get amended into other bills and a bill page can be a year behind the code.
  2. Read the effective-date clause. It is frequently a different date from enactment and is sometimes tied to the next general election.
  3. Check for an injunction. Search the federal district court for the state and the state’s Attorney General as a defendant; an enjoined provision is still printed in the code.
  4. Identify which pattern it is, then find the operative test: for a disclosure statute, the exact required wording and placement; for a prohibition, the intent and knowledge elements and the window.
  5. Cross-check against a maintained tracker — the National Conference of State Legislatures maintains one for AI in elections and campaigns — but treat it as an index into primary sources rather than as the authority. NCSL, artificial intelligence in elections and campaigns.

For the broader state landscape beyond elections, see the US state AI law tracker, and for the provenance-marking approach that operates on the generator rather than on the publisher, see California SB 942 and content provenance.