Nevada Put a Label on AI Political Fakes and Opened a State Investigation Channel for Candidate Threats.

What happens when a Nevada voter watches a video of a candidate confessing to something the candidate never did?

The face looks real. The voice sounds right. The clip arrives two days before voting ends, when a correction may never catch it.

Assembly Bill 73 requires the communication to tell the voter that AI manipulated it.

AB73 covers images, video, and audio intentionally altered with artificial intelligence to create a realistic but false depiction of a person's appearance, speech, or conduct. The result must be convincing enough that a reasonable person would take it as real, and it must leave a fundamentally different impression from the original.

That definition keeps routine editing outside the rule. Cropping a photograph or cleaning up a recording does not automatically turn it into regulated synthetic media. The law is aimed at the AI-generated candidate who appears to say, do, or experience something that never happened.

The required warning is specific. A video must display “This video has been manipulated” for its entire duration. An audio recording must announce the manipulation at the beginning and end. If the recording runs longer than two minutes, the warning must return at least every two minutes.

Picture the clip moving from a campaign account to group texts, reposts, and clipped excerpts. A warning that appears only on the opening frame can disappear along the way. AB73 keeps the disclosure attached to the content so the voter does not have to track down the original upload to learn that the evidence was manufactured.

A candidate shown in an unlabeled fake can ask a district court for an injunction or other equitable relief against the person or committee that made or paid for it. The remedy is designed to stop the communication. The law does not impose liability on internet platforms, broadcasters, cable or satellite operators, or people publishing satire or parody.

AB73 also applies to communications about political or social issues when the speaker intends to influence an election. That brings issue groups and independent advocates into the same legal question as campaigns.

Imagine a church, parent group, or gun-rights organization using AI to illustrate what a proposed policy would look like. If the image realistically depicts a person doing something that never happened, the disclosure may apply even when the communication never says “vote for” or “vote against” a named candidate.

That reach creates the constitutional pressure point. Nevada requires a disclosure and leaves the communication in circulation. The law defines synthetic media around realistic false depictions, protects parody, and limits who can sue. Those boundaries help. The law still compels words on political and issue speech, where the First Amendment is strongest. A smaller organization unsure whether its AI-assisted illustration crosses the line may label everything or avoid the communication altogether.

The voter deserves to know when AI manufactured the evidence. The state must keep the rule tied to realistic deception so it does not become a general warning label for political creativity.

Assembly Bill 123 addresses a different harm. It prohibits a person from knowingly, or with reason to know, communicating a threat or intimidation that expresses an intent to injure a candidate or known family member or damage property. The communication must cause reasonable fear that the threat will be carried out.

Picture a candidate receiving a message that names her child, identifies the family's home, and describes the harm the sender intends to cause. Campaign rhetoric does not explain that away. AB123 gives the candidate a dedicated complaint route and gives the state a criminal law written for that conduct.

A first offense is a misdemeanor. A later offense is a gross misdemeanor. The law expressly protects constitutionally protected speech, including political hyperbole.

The harder issue is the path from complaint to investigation. A candidate may file a written complaint with the Secretary of State. The office must investigate. If it finds reasonable suspicion, it refers the matter to the Attorney General and later reports the outcome to the candidate.

That process can protect a candidate who needs a threat taken seriously. It can also place an election administrator between a politician and a critic. A furious email, an ugly social-media post, or a heated protest sign may trigger a complaint even when the words remain protected. The speaker can face a state inquiry before the line between hyperbole and a true threat has been settled.

AB123's reasonable-fear requirement and express protection for political hyperbole are important safeguards. Enforcement should demand facts supporting every element of the offense. A candidate's fear begins the complaint. It cannot finish the legal analysis.

AB73 tells voters when AI manufactured the political evidence in front of them. AB123 gives candidates and their families a route to report threats that create reasonable fear. Both laws operate beside the First Amendment, where precision carries the whole load.

Nevada should label AI fakes, require proof of a true threat, and keep lawful political dissent out of the government's case file.

This is our read of the laws. We encourage you to read them yourself and reach your own conclusions.