Voluntary AI standards will commence immediately, ahead of a mandatory protection regime

Voluntary AI safeguards will be immediately introduced as the government embarks on four weeks of consultation on mandatory protections aimed at boosting trust in the technology

Industry and Science Minister Ed Husic. Picture: NewsWire / Martin Ollman

Voluntary AI safeguards will be immediately introduced as Labor whittles down the options for regulating safety risks to three key possibilities, including a stand-alone Australian AI act.

Industry Minister Ed Husic will release a departmental paper on Thursday looking at the best way to introduce mandatory protections for the use of AI, saying it is “probably one of the most complex public policy challenges for governments across the world”.

Mr Husic said putting the right regulatory regime in place would “strengthen our firms and … make sure we’re not left behind on an ­increasingly global stage where we’ve got to get our firms competing internationally”.

The departmental paper looks at three regulatory options to boost public confidence and trust in the technology, including the prospect of a New Australian AI act as well as two possible methods of adapting existing regulatory frameworks.

Mr Husic told The Australian he would conduct four weeks of consultation on the best way forward. But he made no commitment on whether the new AI regulatory regime would be introduced before the next election.

“AI can have big benefits for the broader community and economy but we’ve got handbrakes around trust,” he said. “The way in which we bring those guardrails to life … is what’s going to be subject to consultation over the next four weeks.”

Ten mandatory safeguards mitigating harms arising from AI applications are proposed in the departmental paper for inclusion in any new regulatory regime. With no timeline on when the new regime will start, the government has opted to incorporate nine of the safeguards in a new voluntary AI safety standard.

The voluntary standard will commence immediately, starting from Thursday, with the government saying it will help businesses to “ladder up” ahead of the introduction of the mandatory regulatory regime.

Some of the guardrails will require organisations to implement risk-management processes; test and calculate the performance of AI models; enable human control and oversight of the AI systems; inform users when AI technology is being used; and allow those affected by AI systems to challenge outcomes.

The Australian Responsible AI Index 2024 will also be released on Thursday showing just 29 per cent of Australians businesses are currently implementing AI technologies safely and responsibly.

Mr Husic said the index showed “businesses think they’re using it (AI technology) OK, and they may not necessarily be doing that. So to help sidestep that problem, these voluntary standards will help them do just that.”

The departmental paper sketches out some of the new risks posed by AI including its potential to embed human biases in new ways, contribute to harmful deep-fake pornography, and the spread of political disinformation that could disrupt election campaigns.

It noted that in 2023, a Belgian man reportedly ended his own life after a chatbot encouraged him to “sacrifice” himself to stop climate change.

Mr Husic said the government wanted businesses to have the confidence to use AI without burdening them with red tape, but the public expected there would be measures to protect them.

The departmental paper has proposed six principles to help organisations determine if using AI would present a high level of risk.

Organisations would need to make an assessment of the risk of adverse impacts to an individual’s rights; the risk of adverse impacts to an individual’s physical or mental health or safety; the risk of adverse legal effects; the risk of adverse impacts to groups of individuals; the risk of adverse impacts to the broader Australian economy; and the severity and extent of those adverse impacts.