The Australian Electoral Commission is considering requiring deepfakes in electoral communications to be labelled.
In a submission examined by the federal AI inquiry this week, the AEC said consultation on the proposal was underway as part of its review of the instrument governing electoral-authorisation requirements. The proposal has not been adopted. The AEC also emphasised that its authority is confined to that delegated power; any broader regulation of AI-generated political communications would require Parliament to act.
The proposal addresses a significant gap. Existing federal law generally requires regulated electoral communications to identify who authorised them, but AI-generated political content is not automatically unlawful merely because it is synthetic. The AEC warns that realistic material produced and distributed at scale could confuse voters and compromise the actual or perceived integrity of an election—particularly during the short voting period.
Labelling would improve provenance, but it would not resolve every problem. Lawmakers would still need to define what degree of alteration triggers disclosure, allocate responsibility among campaigners and platforms, and account for satire, foreign interference and content generated at negligible cost. Any broader restriction would also require careful treatment of the implied freedom of political communication.
The emerging legal distinction is important: identifying synthetic content is different from regulating whether its political message is true.
Sources: AEC submission; AEC guidance on AI and elections.
Leave a Reply