Australia to Propose Mandatory Guardrails for AI Development

(Bloomberg) — Australia’s government has released options for mandatory guardrails for high-risk research and development of artificial intelligence, including establishing “meaningful” human oversight and ensuring any AI-created content is clearly labeled.

Most Read from Bloomberg

Industry and Science Minister Ed Husic unveiled 10 potential mandatory guidelines for consultation on Thursday in Canberra, while introducing a voluntary safety standard that’s effective immediately. A report by Australia’s Tech Council estimated generative AI could be worth as much as A$115 billion ($77.2 billion) annually to the nation’s economy by 2030.

The regulation of artificial intelligence is one of the “most complex” challenges facing governments around the world, Husic said. “The Australian government is determined that we put in place the measures that provide for the safe and responsible use of artificial intelligence.”

Australia is the latest developed nation to look into controlling the development of artificial intelligence, a technology which is rapidly advancing globally and leading to concerns over the potential consequences for workers, creative industries and the spread of disinformation.

The European Union put in place wide-reaching regulations in March, with the US and UK still weighing their approaches to the technology. The Chinese government has closely monitored the country’s developing AI industry, putting in place 24 guidelines in late 2023.

In January, Husic created a panel of experts to weigh regulations on artificial intelligence, including whether the guardrails should be mandatory or voluntary and what would constitute a “high risk” application of the technology.

The mandatory guardrails suggested by the expert panel this week include:

Enabling human control or intervention in an AI system to achieve meaningful human oversight
Informing end-users regarding AI-enabled decisions, interactions with AI and AI-generated content
Establishing processes for people impacted by AI systems to challenge use or outcomes
Being transparent with other organizations across the AI supply chain about data, models and systems to help them effectively address risks

The 10 restrictions will come into effect on a voluntary basis immediately, with the government to consult on making them mandatory for high risk AI research and development in the future.

Husic said Thursday that the consultation period for the proposed reforms would run until Oct. 4, after which the new guardrails would be legislated.

(Updates with comments from minister.)