Election deepfake

Election deepfakes appear in two distinct forms. The covert kind is a criminal deception — a fabricated robocall or video designed to suppress turnout or smear a candidate, usually timed to the final hours before a vote when there is no time to correct it.

The overt kind is the openly disclosed AI attack ad, where a campaign uses a synthetic version of an opponent as an accepted production technique. Both blur the public's sense of what a candidate actually said.

The honest evidence is that AI deepfakes have not been shown to swing the outcome of a major election; their measured impact is on the information environment rather than on results. The more durable harm is the erosion of a shared record and the "liar's dividend" it feeds.

The practical defence is to treat any unexpected message that discourages or alters how you vote as suspect, and to verify electoral procedure only through official election authorities.