The nation’s top science body has sounded the alarm over major flaws in deepfake detectors, with an international team of researchers calling for urgent improvements as the federal election nears.
The hyper-realistic digitally altered images, video or audio are becoming easier to create and have sparked concerns about the spread of misinformation during the political campaign.
There are no laws governing truth in political advertising, which the crossbench had pushed the government to act on ahead of the election.
Deepfakes were increasingly deceptive, meaning there was an urgent need for detection, CSIRO cybersecurity expert Sharif Abuadbba said.
Deepfakes that “outsmarted detectors” included ones that generated entirely new faces using artificial intelligence, face swaps that make it look like someone else’s face with another in a video and re-enactments that transfer facial expressions and movements from one person’s face to another, the research found.
“As deepfakes grow more convincing, detection must focus on meaning and context rather than appearance alone,” Dr Abuadbba said.
AI is spurring the creation of deepfakes and magnifying their reach and impact.
Australia signed onto a global statement of inclusive and sustainable AI following an action summit in Paris in February centred on the ethical use of the technology.
“We believe AI can be an opportunity to bring extra tools to fight against disinformation and information manipulation rather than used maliciously as part of information manipulation campaigns,” French ambassador in Australia Pierre-Andre Imbert told AAP, calling for more collaboration between the two nations.
Home Affairs Department deputy secretary Nathan Smyth warned democracies were being undermined by the spread of disinformation, including by foreign actors seeking to undermine democratic processes.
The Australian Electoral Commission has started a campaign on TikTok – which Mr Smyth branded as influential in elections – to reach young voters and tackle misinformation by putting forward tools to spot fake posts.
The federal government is working on mandatory guardrails for AI and high-risk systems.
The Home Affairs Department identified foreign interference as a major risk to democracy and said it was being used to sway elections and spread disinformation to reduce voter confidence in electoral processes.