S. 1213: Protect Elections from Deceptive AI Act

S. 1213, the Protect Elections from Deceptive AI Act, makes it illegal to knowingly spread AI-faked audio or video of a federal candidate when the goal is to swing an election or raise money. It's narrow on purpose: the target isn't all synthetic media, just deceptive AI content aimed at federal politics.

The bill spells out what counts. The media has to be built by AI — machine learning, deep learning, and similar techniques — that either alters real footage or conjures something entirely fake that still looks authentic. Then it applies a reasonable-person test: would an average viewer come away with a fundamentally different impression of what the candidate looked like, said, or did? If yes, it can qualify.

Enforcement runs through the courts, not prosecutors. A candidate who's the subject of the media can ask a judge for an order to halt distribution, and those requests get bumped to the front of the line. The candidate can also sue for damages and recover attorney's fees if they win. But the bar is steep — they have to prove the violation by clear and convincing evidence, a tougher standard than most civil cases.

The bill carves out room for journalism and comedy. News outlets — broadcast, cable, streaming, newspapers, and online publications — are exempt when they show the media as part of real news coverage and clearly flag that its authenticity is in question. Satire and parody are exempt outright. The likely fights: what counts as "deceptive," how clear a disclosure has to be, and whether courts can move fast enough to matter during a campaign.