Australia introduces AI voluntary standards to improve safety

Australians will be protected from high risk uses of artificial intelligence like in hiring and firing, facial recognition in public or disinformation under sweeping new safety measures.

Australians will be protected from high risk uses of artificial intelligence like in hiring and firing, facial recognition in public or disinformation under sweeping new safety measures.

A new voluntary AI standard will unveiled by the federal government on Thursday, which is designed to increase trust in the emerging technology, which is estimated to contribute as much as $115 billion to the Australian economy by 2030.

Industry and Science Minister Ed Husic said Australians understood AI could be useful, but are “genuinely concerned about the risk of the technology going wrong“.

“We’ve heard that loud and clear,” he said.

“We think the work that we are releasing can help build trust, but also send a serious signal to people that don’t use the technology (safely) that there will be consequences.”

Industry and Science Minister Ed Husic will unveil the voluntary standards for AI on Thursday. Picture: Brendon Thorne/Getty Images

Mr Husic said if people didn’t trust AI they wouldn’t use it, and then businesses would not “benefit” from the technological advances.

“The risks in terms of the use of AI are serious, for example if AI is used to hire and fire people, the way AI is used with kids, the use of facial recognition in public places, these are all serious matters that we think need to be dealt with,” he said.

Australia will consult on mandatory guardrails for AI. Picture: Olivier Morin/AFP

Labor will also release ten new mandatory guardrails, which will be put out for a short four-week consultation ahead of their implementation.

These include requirements for companies to test AI models, produce risk management plans and require some level of human oversight for “high risk” models.

Mr Husic said while work on the mandatory measures was ongoing, he wanted businesses to “seize the moment” and adopt the voluntary standard.

“(This) can show them how to better deal with the risks associated with AI, and we encourage them strongly to follow that standard,” he said.

Generative AI is being increasingly used by businesses in Australia. Picture: Marco Bertorello/AFP

The sweeping reforms could lead to the introduction of an Australian AI Act to regulate and ensure compliance.

A report released alongside the measures found bias in AI can be due to inaccurate and insufficient data, or because the technology is poorly designed or deployed.

It also said harms to people included breaches of privacy, exclusion from access to opportunities and services and discrimination.

AI could also harm society more broadly through misinformation and disinformation, eroding social cohesion or spreading extreme or misogynistic content.

“Deepfake” images, video and audio generated by AI are also becoming increasingly difficult to detect as the technology becomes more sophisticated.

Generative AI alone has been estimated to contribute between $45bn to $115bn to the Australian economy by 2030.

However only 29 per cent of businesses are implementing AI “safely and responsibly,” according to new Responsible AI Index data.