Debating faked images and video, Alaska lawmakers say AI could eventually be a person

The Alaska House of Representatives is moving toward new restrictions on the use of artificial intelligence computer software that can be used to impersonate a person’s voice and appearance.

As soon as Thursday, the House may vote on House Bill 358, which would ban someone from using a faked picture, video or sound clip to influence an election, unless the fake, known as a “deepfake,” is accompanied by a disclaimer.

On Wednesday, lawmakers debated amendments to the proposal and rejected, by a 16-24 vote, an outright ban on deepfakes in elections.

Rep. Justin Ruffridge, R-Soldotna, suggested the ban and said that by definition, a deepfake is intended to “trick people,” and therefore should not be allowed.d

Rep. Mike Cronk, R-Tok/Northway, sponsored the bill.

The committee held eight hearings on AI software this year, she said, and lawmakers need to balance the First Amendment with their desire for regulating brand-new software.

The committee used a “national industry standard” of AI, she said, and while AI deepfake issues have not been tried in the courts, “meaning it could go either way,” a ban “does raise that First Amendment concern a little bit more” than the current bill’s disclaimer requirement.

Rep. Sarah Vance, R-Homer, said she believes that if a candidate were to use a deepfake of his or her opponent, it would backfire.

Rep. David Eastman, R-Wasilla, proposed six amendments to the bill, and all failed, but not before sparking a debate among members of the House about whether artificial intelligence should be defined as a person or not.

Eastman suggested that the bill be amended to exclude AI from the definition of a person.

Lawmakers overwhelmingly voted down that suggestion, with Eastman being the only member of the 40-person House to vote against AI personhood.

Rep. Jesse Sumner, R-Anchorage, said AI development could eventually lead to an artificial intelligence able to reason and make decisions independently of human interaction. 

Sumner, semi-kidding, said that if a truly independent artificial intelligence is developed, he likes to think “it would read the statute and … follow the law.”

Rep. Jamie Allard, R-Eagle River and a member of the judiciary committee, spoke in response to one of Eastman’s early amendments and said that if lawmakers — or the general public — are unhappy with the current legal restrictions on AI, they should stay involved.

“We’re going to be here next year. We’re going to keep doing this until we can keep up with the artificial intelligence that’s ever-changing,” she said.

Correction: The bill’s sponsor and Vance’s role on the House Judiciary Committee were misidentified in the original version of this article.