A proposal before the Maine Legislature aims to address the growing use of "deepfakes" in political campaigns.
Concern is increasing in Maine and nationally about the use of artificial intelligence, or AI, to create deceptive videos, audio or other political messaging that could influence the outcome of elections. For instance, some voters in New Hampshire received a robocall last year from an AI-generated voice that sounded like former President Biden urging them not to vote in the state's primary election.
Two dozen states, to date, have passed laws aimed at regulating AI in political communications in some form, ranging from outright bans on deepfakes to measures aimed at increasing transparency.
A bill under consideration in the Maine Legislature would not prohibit political operatives from using AI to create "deceptive of fraudulent political communication." But the bill, LD 1690, would require a disclosure statement saying "This communication contains audio, video and/or images that have been manipulated or altered."
Bill sponsor Rep. Amy Kuhn, D-Falmouth, said the bans and criminal penalties enacted in some states have raised constitutional questions about limiting political speech.
"This approach adopts the viewpoint that the remedy to disagreeable speech is more speech," Kuhn told members of the Legislature's Veterans and Legal Affairs Committee. "Adding more information to the marketplace of ideas allows viewers and voters to consider the content with eyes wide open."
The disclosure requirement would not apply to political "satire or parody." But Kuhn said failure to address the issue could result in voter confusion and a loss of confidence in elections.
The bill was endorsed Monday by Public Citizen, a national progressive advocacy group that has worked on AI-generated political messaging at the national and state level.
Ilana Beller, who leads Public Citizen's state legislative work on AI, said many of the laws adopted in other states have enjoyed strong bipartisan support because deepfakes can be used against any political candidate.
"It is worth noting that recent advances in the realm of AI have made it such that anyone can generate a deepfake for either no cost or very low cost," Beller said. "The technology is getting really, really good. Unfortunately, recent studies have shown that the majority of Americans at this point cannot determine what is a deepfake video versus what is real content when it's a high-quality deepfake."
Failure to disclose the use of AI in deceptive or fraudulent political communication could result in fines of up to 500% of the cost of the ad or political messaging under Kuhn's current proposal.
Representatives for cable and internet service providers as well as the Maine Association of Broadcasters asked lawmakers to make it clearer that their clients would not be held accountable if an ad or other communication does not include the necessary disclosure. Instead, they said the penalty should fall on the creator of the content.
Kuhn said she is working with the groups and will introduce additional language to address those concerns and others raised by the Maine attorney general's office. No one testified against the bill during Monday's public hearing.
The committee is slated to hold a work session on the bill on Friday.
