Deep Fakes in Elections (S.23) - Overview & Analysis

The bill aims to regulate the use of synthetic media (deepfakes) in elections by requiring clear labeling of realistic, deceptive AI-generated content in the 90 days before an election and giving prosecutors and candidates new tools to respond to harmful deepfakes.

The Details:

• Synthetic Media must include a disclosure stating: “This media has been manipulated or generated by digital technology and depicts speech or conduct that did not occur.”

  • For images and video, the text must be clearly readable (and accessible to people with disabilities)

  • For video it must appear for the full duration.

  • For audio, the disclosure must be clearly spoken at the beginning and end, and at least every two minutes for longer recordings.

• These rules only apply within 90 days before an election in Vermont and only when the person sharing the content knows it is deceptive and fraudulent synthetic media.
  
  
• The bill defines “Synthetic Media” as images, audio, or video of a person’s appearance, speech, or conduct that are created or intentionally manipulated using digital technology, including artificial intelligence.

• Additionally, the bill defines “Deceptive and fraudulent synthetic media” as synthetic media that looks realistic to a reasonable person and either:

  • Depicts a political candidate in a way that injures the candidate’s reputation, or

  • Depicts any individual who is trying to unduly influence the outcome of an election by providing voters with materially false information.

• A “reasonable person” standard is used to determine whether the content appears realistic and deceptive, which is intended to exclude obviously fake or joking content.

• The bill creates several key exceptions.

  • Bona fide news programs and commentary that clearly flag doubts about the authenticity of the media are exempt.

  • Broadcasters and platforms are also exempt when they are paid to air the content or required by federal law to carry candidate ads.

  • News websites and periodicals that routinely publish news and opinion are exempt if they clearly state the content does not accurately represent the person’s speech or conduct.

  • Satire and parody are explicitly exempt, as are telecom and online service providers and interactive computer services when they are transmitting or hosting content created by someone else.

• Penalties for knowingly and intentionally violating the law range from fines of up to $1,000 for basic violations, up to $5,000 when done with intent to cause violence or bodily harm, up to $10,000 for repeat offenders, and up to $15,000 when both prior violations and an intent to cause harm are present. Obstructing an investigation (for example, by destroying or falsifying evidence) carries an additional fine of up to $5,000.

• Candidates whose appearance, speech, conduct, or environment is misrepresented through deceptive and fraudulent synthetic media can seek injunctive or other equitable relief in court to stop the distribution. The Attorney General and State’s Attorneys are also given broader enforcement powers for election-related violations, including civil investigative tools to request documents, sworn answers, and testimony.

• The bill would take effect upon passage.

Analysis:

Twenty-one states have passed similar legislation and there is a growing recognition of the need to address the risks posed by deep fakes in elections. Protecting the integrity of elections and the democratic process is critical to maintaining our democracy and new AI-powered tools are making it difficult to determine what is real and what is fake.

One notable incident in Slovakia occurred where an audio deep fake circulated just before an election, suggesting a candidate wanted to rig the vote. The use of deep fakes in campaigns is concerning and almost certainly poses a risk to free and fair elections.

Other states have faced free speech challenges over similar pieces of legislation. Lawmakers will need to be careful when crafting the language to not violate free speech rights. This is one of the reasons why the bill only requires the labeling of deep fakes instead of outright banning the practice in elections.

Additionally, there is a risk that citizens or campaigns may inadvertently violate this law by sharing content on social media that they didn't realize was AI-generated. Revisions to the bill do offer some protections for persons who share content they didn't know was AI generated, but it is a "reasonable person" standard that may not be fully inclusive.

However, these are all navigable problems and some protection for campaigns against deep fake attacks would be welcome.

Current Status:

The bill was signed by the Governor on March 5, 2026.

Last updated: 3/6/2026

DISCLAIMER: Generative AI used to assist in the production of this report.