Senate committee debates S.23 requiring labels on AI 'deepfakes' near elections

MONTPELIER — The Senate Government Operations Committee on Thursday reviewed S.23, legislation that would compel disclosure of AI-generated “synthetic” media used to depict political candidates or influence voters in the period before an election.

Representative Shay Waters Evans, the ranking House sponsor, told the Senate the House edits were designed to clarify — not weaken — the bill and to make enforcement more straightforward for the attorney general’s office. “We don’t wanna stifle free speech. We just wanna make sure that people are aware of what they’re consuming,” Waters Evans said as she summarized changes that narrow the definition to media representing political candidates and add distribution to the bill’s scope.

The committee spent much of its discussion on three core tensions: how broadly to define “synthetic media,” whether liability should require actual knowledge or a reasonable-person “should have known” standard, and how the bill should treat distributors and sharers. Rick Seigel of the office of legislative counsel cautioned that the Senate penalty language used a knowing standard, and widening the mental-state requirement to a “should have known” reasonable-person test…