Australians are set to be bombarded with AI-powered disinformation ahead of the next election, the electoral commission has warned amid calls for a ban on "deepfakes" that could deceive voters.
AEC Commissioner Tom Rogers said the commission was largely powerless to stop a repeat of the AI-powered misinformation that plagued recent elections in the United States, South Korea and India.
"Our legislative toolkit is very constrained in what we can deal with and not deal with," Mr Rogers told the first hearing of a Senate inquiry on adopting artificial intelligence on Monday.
"Of course I would prefer that there was no misinformation but, under the legislation, if it's authorised, it's currently lawful and that would be a matter for Parliament."
In the lead-up to the US New Hampshire presidential primary in January, an AI deepfake robocall, reported to have used voice cloning technology impersonating US President Joe Biden, urged voters to skip the primary election.
Mr Rogers was asked if the AEC would be able to crack down on "a barrage of robo-calls" in marginal seats on the eve of the election from someone purporting to be, and "indistinguishable from Peter Dutton", making "false, malicious" statements that blatantly misrepresented the Coalition's policy positions. He said such a call would not breach the Electoral Act if it was duly authorised.
Independent Senator for Canberra David Pocock and Curtin MP Kate Chaney have written to Trade Minister Don Farrell and Industry and Science Minister Ed Husic demanding a ban on the use of generative AI in pre-election campaigning.
"In the absence of any federal truth in political advertising laws or a regulatory framework for the use of AI, Australia is acutely vulnerable to the misuse of this evolving technology to undermine our democracy," they wrote. "Noting the recent establishment of the Artificial Intelligence Expert Group, we ask that the government tasks this body with an urgent examination of the steps and reforms needed to update Australia's electoral laws to prohibit politicians, candidates and relevant third parties from deliberately spreading misinformation during election campaigns."
The use of any AI-generated image, audio or other material during an election campaign should be prohibited, they wrote. "Such changes need to be brought before the Parliament and implemented ahead of the next federal election being called, which we note could be held as early as August this year."
In a written submission to the inquiry, the commission said: "If a person seeks to disrupt a federal electoral event by using AI, the AEC only has limited powers to investigate or take action."
The Commonwealth Electoral Act makes it an offence to publish materials likely to mislead or deceive an elector in relation to the casting of a vote, but the courts have interpreted this in a restrictive way that concerns the mechanics of voting.
Mr Rogers said mandatory or voluntary digital watermarking of materials produced with generative AI would help the commission to detect any materials that may breach the act.
He also called for a national digital literacy campaign to help voters better understand how to distinguish misinformation, including that generated with AI.
