Australian Gov't Proposes Mandatory 'Guardrails' For Use Of AI

CANBERRA, Sept 5 (Bernama-Xinhua) -- The Australian government has flagged plans to introduce new laws to regulate the use of artificial intelligence (AI).

Ed Husic, the minister for industry and science, on Thursday released a discussion paper proposing 10 "mandatory guardrails" to regulate AI.

The proposed guardrails include human oversight and the ability to challenge the outcomes of automated decision-making.

Husic described AI as a great opportunity for Australia but said that current standards for its use are insufficient.

"Australians know AI can do great things but people want to know there are protections in place if things go off the rails," Husic said in a statement.

"From today, we're starting to put those protections in place."

The government has proposed three possible options to set rules for AI -- adopting the guardrails within existing regulations as needed, introducing new legislation to adapt existing framework, or a new standalone cross-economy AI act, reported Xinhua.

The standalone act would enable the establishment of an independent regulator for AI, though the research paper said that setting up the new body would take resources and time.

Thursday's discussion paper proposes to define high-risk AI in two broad categories -- the first relating to instances where proposed uses of an AI system are known or foreseeable and the second relating to highly capable models where all possible risks cannot be foreseen.

It said that AI's potential to cause harm includes creating and amplifying bias, physical or psychological injury, breach of privacy and threatening national security.

According to the Tech Council of Australia, the peak industry body for the technology sector, generative AI could contribute up to 115 billion Australian dollars (US$77.2 billion) per year to the country's economy by 2030.

Husic also on Thursday released a new voluntary set of AI safety standards to provide practical guidance to businesses to implement best practices in the use of AI. 

-- BERNAMA-XINHUA