Australian political parties are embracing what has been described as “appalling and disgusting” deep fakes and other generative AI, with few regulatory or legal barriers stopping their use to influence elections.

Queensland Premier Steven Miles called the use of generative AI and deep fake content “appalling” after the Queensland Liberal-Nationals Party posted a video of the premier dancing on TikTok with a caption that criticised him for posting videos on the platform during a cost-of-living crisis.

His federal colleagues who run the Australian Labor Party’s TikTok page had previously posted a video of Peter Dutton in the same format in June.

Professor Toby Walsh, chief scientist at the University of New South Wales’s AI Institute, said there needs to be legal and regulatory barriers to the use and dissemination of AI-generated content.

“The technology is evolving very rapidly and unless you are in the room and see it with your own eyes, you can’t say for sure if it is synthetic or not,” he said.

“One hundred-plus years ago when we invented photography, we invented fake photography at the same time, but the difference is the technology has been very democratised. It used to be you needed some skill and sophistication to access it.”

Miles said: “It’s appalling and disgusting that David Crisafulli has stooped to using AI and deep fake videos to attack me … In a time when misinformation is everywhere, we as politicians have a duty to communicate with our audiences and to voters clearly and honestly.”

Electoral impact

In May, a parliamentary committee heard that the Australian Electoral Commission (AEC) did not believe it had the legislative tools or internal technical capability to handle false AI-generated images during elections.

AEC commissioner Tom Rogers told the committee that voice-cloned robocalls – like those that resulted in a man being fined $6 million in America – wouldn’t necessarily be illegal under current laws.

The same technology is being used in Australia, with Sean O’Reilly, a Dandenong councillor, creating rap songs in languages he doesn’t speak to appeal to Vietnamese, Chinese and Khmer constituents.

Evan Ekin-Smyth, the AEC’s national spokesperson, said this will be the first federal election in Australia where AI is potentially used as a part of campaigning or political commentary in a prominent way.

“Elections internationally in recent times have seen it in different ways, be it deceased former President Suharto appearing in a deep fake during the Indonesian election or Eminem apparently endorsing a candidate in South Africa,” he said.

“To think Australian elections will be immune to that potential is perhaps naive.”

He said the rise of AI provides societal opportunities and challenges.

“It’s important not to overstate the danger of AI because we know that can undermine public trust in all forms of communication,” Ekin-Smyth said.

“AI is a force multiplier for electoral disinformation, but it’s still only as useful as the individuals behind the keyboard.”

Social responsibility

Zoe Hawkins, head of policy design at ANU’s Tech Policy Design Centre, suggested a requirement to label AI-generated content when it is used in advertising and politics would be a good place to start.

“The way the public debates is important. We need to make sure we don’t erode trust in democracy as a self-fulfilling prophecy,” she said.

“We have to be mindful about how we discuss this policy challenge. It’s about keeping Australians confident in the integrity of the process.”

Although the use of artificial intelligence wasn’t common in the British general election, the Reform Party was falsely accused of running AI-generated candidates.

Walsh agreed with Hawkins’ suggestion and also added that more needs to be done to hold social media companies accountable.

“We saw what happened with Cambridge Analytica, and what worries me is that should have been a wake-up call,” Walsh said.

“Social media was a cheap, effective way to reach into people’s lives and we’re now about to supercharge that by adding persuasive AI-generated fake content.”