AI, deepfakes and political campaigns: mass lawmakers consider rules

Sen. Barry Finegold, D-Andover, seen in an April photo, proposed a bill to regulate the use of AI in campaign materials.

BOSTON — Concerns about the possible use of computer-generated images and sound recordings that could confuse and alienate voters prompted Massachusetts lawmakers to consider a bill Tuesday that would require all materials AI-generated campaign campaign carries a warning label.

The bill, proposed by Sen. Barry Finegold, D-Andover, and Rep. Frank Moran, D-Lawrence, would ban the use of synthetic media in the 90 days before an election without disclosing that the material was manipulated or generated by artificial means. intelligence.

The bill would also prohibit the deliberate manipulation of records to confuse or deceive voters by presenting a candidate or political party in a manner contrary to their stated programs and policies.

The legislation also provides consequences for violators: the possibility of filing a civil action, penalties of up to $10,000 per incident and “any other relief that the court deems appropriate.”

Tech companies back their bill, want protections for AI creators

Christopher Gilrein, executive director of TechNet, the organization that bills itself as the voice of technological innovation, said businesses fully support the bill.

“But we want it to look like the bills passed in other states that have filed similar legislation,” Gilrein said.

Speaking before the Joint Committee on Election Laws, Gilrein asked that responsibility be placed solely on the creator of the material, not on the tools and programs used, nor on the platforms that distribute the material.

“Artificial intelligence has the potential to solve our most pressing problems. We support clear disclosure when AI is used in communications,” Gilrein said. He also requested that the bill contain permissions for programmers to investigate the material to determine who created it and what tools were used.

Gilrein cited the recent use of a deepfake in the New Hampshire primary: a robocall using a facsimile of President Joe Biden’s voice, which falsely suggested that if voters cast ballots in the January primaries, they could be prevented from voting in November. The call was made on a Sunday evening, just two days before Tuesday’s primaries.

Biden was not on the ballot in New Hampshire but won the primary based on write-in votes. Investigators learned the fake call was created by a political consultant who claimed it was a wake-up call about the proliferation of AI-generated campaign material and the potential for its misuse.

Using deepfakes to influence voter choice could lead to a loss of confidence in the US electoral system, according to Hamid Ekbia, a professor at Syracuse University, speaking for the Academic Alliance for AI Policy, a coalition of 40 major universities. The social and political implications, Ekbia said, include increased polarization and negative partisanship across the political spectrum.

13 states have already passed laws

According to Public Citizen, a national advocacy organization, 13 states have already passed laws regulating AI.

“The federal government has been slow to act,” said Craig Holman, the organization’s lobbyist. “It’s up to states to step up their efforts.”

He described the 2024 election cycle as the first where AI could have a profound impact on campaigns and voter choice.

“It’s almost impossible to tell the difference between what’s real and what’s been computer-generated,” Holman said.

He discussed the 2023 Chicago mayoral campaign, in which a widely circulated deepfake appeared to record a candidate condoning police brutality. The candidate, Paul Vallas, ultimately finished second after a close race.

The bill provides protections for traditional and social media platforms. A media outlet would be allowed to use manipulated material in a story as long as it clearly states that the material is false.

Additionally, media outlets could not be prosecuted if they make efforts to verify the veracity of material before using it in a report or broadcast. If they knowingly distribute false material, they must visibly alert the public that it does not reflect the speech or behavior of the candidate depicted.