Industry and Science Minister Ed Husic has told a Newcastle conference on artificial intelligence that the government is "playing catch-up" on regulating the rapidly evolving technology.
Mr Husic was one of 17 speakers at a Business Hunter summit who examined the potential benefits and dangers of AI under the banner "AI: Friend or Foe?".
Many governments are scrambling to develop regulations around the technology, which has progressed in recent months to creating text, images, audio and video virtually indistinguishable from human-generated content.
Mr Husic said there was "merit" in proposals for mandatory labelling of content produced by AI software like ChatGPT and image generator Midjourney.
"It is a serious issue that we need people to have confidence that the stuff they see on their screens ... is legit," he said.
"It was expected that it would take a lot longer for AI to be able to generate image, and the way it's happening now is a serious issue.
"The broader public have to have confidence that what you're seeing is real."
He said the government had started consultations about labelling and other AI-related regulatory issues but lawmakers had not kept pace with the development of AI.
"I'm not going to sugarcoat it: I think we absolutely have been behind. We're playing catch-up," he said.
"And I'm hoping this work starts to prompt in people's minds across government that we've got to get ahead on it."
He pointed to American politics as a warning of how "fake news" could be harnessed to hijack democracy.
"Disinformation as we see it play out in the 2016 US presidential election is serious stuff," he said.
"We don't need that to take hold more and more."
Australia's Human Rights Commissioner Lorraine Finlay told the conference via video link that generative AI was a "game-changer" which would make "massive disinformation" campaigns cheaper and easier.
"The use of deep fakes means distinguishing between what is fact and what is fiction becomes increasingly difficult," she said.
"This poses a particular challenge for countries like Australia, for democracies, because we rely on our citizens having access to truthful information."
She said regulation needed to be "durable enough to be able to accommodate developments as they come".
Ms Finlay warned of facial recognition technology being used in law enforcement and algorithmic bias undermining the rights of certain groups in society, including in job interviews.
She called on the government to develop "guardrails" to protect the public, including appointing an AI safety commissioner.
"The law is really struggling to catch up. While I think the law reform side of things is incredibly important, we can't wait for the laws to be perfect. We need to start acting now."
The conference, hosted by maths and science broadcaster Adam Spencer, heard from AI experts at Microsoft, CommBank, CSIRO, nib Group and engineering firm GHD.
Mr Spencer outlined the power of generative AI and traced the rapid trajectory of AI performance since chess champion Garry Kasparov beat IBM supercomputer Deep Blue in 1996.
Maitland-born Microsoft business applications technical specialist Brent O'Toole pitched the company's suite of natural language-prompted AI software being built in to its established products.
Mr Husic said Australian governments could use AI in "some of the decision-making processes just to speed things up", despite the Robodebt scandal plaguing the previous Coalition regime.
"I can see a scenario where there that emerges in due course. We're not there yet. The confidence in the way the government's managing the technology has got to lift before we get to that point," Mr Husic said.
