Scare campaigns: Victoria’s major parties trade blows over AI-generated videos of leaders in election ads

Victorian political parties used AI-generated videos to make opponents appear to say or do things they had not. The tactic mattered because the videos looked and sounded convincing while presenting misleading political claims. Experts warned that such content could further weaken trust in politicians and democracy.

Labor posted a video depicting Liberal leader Jess Wilson holding bills for a $150,000 heart attack, a $6,000 ambulance and a $40,000 birth. Wilson never made those statements. The Liberal campaign then used AI to depict Premier Ben Carroll asleep at his desk, following allegations in the Herald Sun that he napped at work.

The attacks became reciprocal. A Labor source said the Liberals had “thrown the first brick”, while the Liberal party called Labor’s video “garbage”. Experts and the Greens called for a ban. The article also reports calls for truth-in-advertising laws and clear AI labelling.

An AI-generated deepfake uses artificial intelligence to create or alter a person’s image, voice or actions. It can make someone appear to say words or perform acts they never said or performed. The danger is not only that the content is false, but that it can look and sound authentic enough to influence voters.

Labor’s Halloween video showed Jess Wilson clutching a bill and apparently warning about healthcare costs. It used a depiction and voice that sounded like Wilson, although she never made those claims. The Liberal campaign created a video showing Ben Carroll asleep at his desk, based on allegations about his work habits rather than verified footage.

The article says AI makes subtle mischaracterisations easier to represent artificially. Only the Advance Australia video carried an AI disclaimer. Experts therefore urged political parties to use watermarks or labels, while South Australia requires labels on AI election advertisements.

Several political parties and groups used AI-generated material during Victoria’s election campaign. The major parties used it against each other, while conservative groups used it for broader political messaging. Their videos ranged from fabricated statements and behaviour to imagined election scenes and inflammatory imagery.

Victorian Labor posted a video making Jess Wilson appear to promote expensive American-style healthcare. The Liberal party posted an AI video showing Ben Carroll asleep at his desk. Advance Australia released a video imagining a Labor election victory, featuring former premiers Jacinta Allan and Daniel Andrews. One Nation showed a video at its launch that invoked inflammatory tropes about people from different ethnic backgrounds.

The Liberal campaign called Labor’s video “garbage”, even as it used AI itself. A Labor source said the Liberals had started the exchange. Transparency International Australia, experts and the Greens warned that these tactics could damage public trust and prompted calls for bans or stronger labelling.

AI-generated political content was not an isolated incident in this campaign. The article describes at least four separate video examples involving Labor, the Liberals, Advance Australia and One Nation. It also says political misinformation predates AI, but synthetic media makes misleading messages more convincing and easier to produce.

The examples are Labor’s video portraying Jess Wilson and her healthcare claims; the Liberal video showing Ben Carroll asleep; Advance Australia’s fabricated Labor election-night victory; and a One Nation launch video using inflammatory ethnic tropes. The first two were direct attacks by major parties. The latter two came from conservative political groups.

The article does not provide a complete count of every AI post, so four is the number of described examples, not necessarily the campaign’s total. Only Advance Australia’s video included a disclaimer. Experts and the Greens called for bans, while Clancy Moore urged parties at least to use AI watermarks.

Misleading political videos can damage trust because they give false claims the appearance of direct evidence. Viewers may see a familiar politician’s face and hear a familiar voice, then assume the person genuinely made the statement. That makes synthetic content more powerful than an ordinary written allegation.

The Labor video made Jess Wilson appear to endorse costly American-style healthcare, including a $150,000 heart attack treatment. Wilson never said those words. The Liberal video made Ben Carroll appear asleep at his desk. These examples used realistic-looking scenes to strengthen arguments that the article says did not hold up.

Clancy Moore said trust was already at an all-time low and that lies, deepfakes and AI-generated politicians could further erode confidence in politicians and democracy. Timothy Koskie said even small misstatements need follow-up. The article reports calls for truth-in-advertising laws, penalties and clear AI labels.

South Australia’s truth-in-political-advertising laws target both deception and disclosure. They ban AI-generated deepfake electoral ads that show a person performing acts they did not perform, unless that person gave permission. The laws also require advertisements using AI to be labelled, so voters can recognise synthetic content.

A person may consent to an image being used in a deepfake advertisement. Even then, the advertisement must carry a label if it uses AI. Breaches can attract penalties of up to $5,000. These rules directly address the kind of fabricated politician videos described in Victoria’s campaign.

The article says equivalent truth-in-political-advertising laws exist in the ACT and South Australia, while Clancy Moore is calling for Victoria and the federal government to introduce them. Victoria’s current situation is presented as lacking those protections. Only Advance Australia’s video had an AI disclaimer, and experts urged watermarking at minimum.

Protecting elections is difficult because misleading political messaging already existed before AI, and social media makes it easy to circulate. AI adds convincing faces, voices and scenes. This lets campaigns turn a subtle mischaracterisation into a video that looks like evidence, before fact-checkers or regulators can respond.

Labor’s video made Jess Wilson appear to deliver healthcare warnings she never gave. The Liberals answered with a video of Ben Carroll apparently asleep. Timothy Koskie said the issue was not necessarily tricking voters into believing an entirely false story, but making an argument that does not hold up seem more convincing.

Koskie said stopping the practice is very difficult and the public may not reward those who avoid it. He argued that misleading actors should face follow-up, including small fines. The article reports calls for truth-in-advertising laws, AI watermarks and penalties, but says only South Australia and the ACT currently have such laws.