Misleading media made using artificial intelligence within 90 days of an election would have to include clear disclosures under a bill moving through the legislature.
If it becomes law, it would mark the first time the state regulates AI in statute, said Rep. Matt Birong (D-Vergennes).
“It’s about maintaining, within the media and social media landscape, the integrity of the information that’s being disseminated and put out to the public,” Birong said.
The bill gained traction last session in response to an illegal robocall campaign used ahead of the 2024 primary in New Hampshire. The calls used an AI-generated “deepfake” voice made to sound like then-president Joe Biden that urged voters to skip the primary.
The prospect that AI-generated audio, images and videos could be used to influence elections is a major concern of those focused on election security and integrity. President Donald Trump and his administration have shared AI-generated content, heightening fears it will be commonplace.
In recent weeks the White House shared a digitally altered image of the arrest in Minneapolis of attorney and activist Nekima Levy Armstrong. The original image showed her composed as she was led away by a federal agent. The White House released a manipulated image that showed Armstrong hysterically crying and with darker skin.
A racist video posted to Trump’s Truth Social account last week depicted Barack And Michelle Obama as apes. The White House initially defended the clip as an “internet meme,” but it was removed after even Republicans expressed outrage.
The challenge in drafting the bill was to find a way to protect the integrity of election information without violating people’s free speech rights, Birong said. This was accomplished largely not by restricting people’s free speech, but by requiring the disclosure of the method used to create the content, he said
The bill says people cannot within 90 days of an election “publish, communicate, or otherwise distribute” any “synthetic media” that they know is “deceptive or fraudulent” unless the material contains a disclosure clearly stating: “This media has been manipulated or generated by digital technology and depicts speech or conduct that did not occur.”
For audio recordings, such as radio ads, the same disclosure must be read in a clear voice at the beginning and end of the recording, and in the middle of recordings longer than 2 minutes.
The bill exempts media “constituting satire or parody.” Also exempted are media outlets such as newspapers and magazines that publish or report on such media “if the publication clearly states that the deceptive and fraudulent synthetic media does not accurately represent the speech or conduct of the represented individual.”
The bill includes $1,000 penalties for a first offense, or up to $15,000 for multiple offenses and the posts or memes incite violence. The Attorney General’s Office or State’s Attorney could investigate violations, and a person harmed by such media could seek relief through civil courts.
