High-Risk AI Applications Could Face Bans Under Potential Law Reforms

Australians are increasingly concerned by the intrusion of artificial intelligence (AI) into daily life, and the government is working on ways to constrain it, Labor MP Patrick Gorman has told the Australian Lawyers Alliance national conference in Fremantle.

Gorman is Assistant Minister to the Prime Minister and the Attorney General, as well as Assistant Minister for the Public Service.

Previewing a forthcoming Susan McKinnon Foundation report on public attitudes to AI, he said the one point all 3,000 people surveyed agreed upon was that they opposed what he called "the AI intrusion."

The "left and progressives" gave the issue a negative 15 percent rating, while the "right and conservatives" were even more concerned, scoring it at negative 20 percent.

In this fast-changing environment, Gorman said the role of lawmakers is to "balance risk with opportunity," and "to shield the Australian public from the dangers of AI while not restricting the potential for AI to deliver positive and profound improvements in living standards."

"We know the risks of having our sensitive data harvested and used," he said.

"Your information could be training AI without your knowledge or consent. AI generative content can also be created to mimic the works of existing Australian artists and creatives." He also flagged AI's potential to exacerbate issues like cybercrime.

One possible legislative model under consideration is the European Union Artificial Intelligence Act, which assigns AI systems and applications to one of three risk categories: unacceptable risk, high risk, and minimal risk.

In this framework, unacceptable risk systems and applications are prohibited. It is this model which Canada has broadly followed, Gorman said, and seems to be the direction in which other countries are headed.

"Last year in the UK, an AI white paper was released which argues for the risk-based approach," he told the lawyers.

"The paper classifies AI systems based on the level of risk they pose. It emphasises the development of AI systems that are human-centric and trustworthy whilst also promoting innovation through the development of AI innovation hubs to support research and development."

In the United States, the first state-based AI legislation has been passed. Known as the Colorado AI Act, it will come into effect from February 2026, requiring developers of high-risk artificial intelligence systems to use reasonable care to protect consumers from foreseeable risks of algorithmic discrimination.

"These are all developments that the Australian Public Service is monitoring closely," Gorman said.

Ed Husic, Minister for Industry and Science, recently held consultations on proposals to introduce mandatory guardrails for AI in high-risk settings, a process that is now helping shape the government's AI policy.

The Senate Select Committee on Adopting AI is also investigating the technology's opportunities and impacts and is due to present its final report on Nov. 26.

The Attorney-General's Department is currently looking into AI-related law reform across several areas, including privacy, copyright, automated decision-making, cybercrime, and technology-facilitated abuse.

He added that the government was working on policy that would ensure that systems and processes are "sufficiently robust [and that] flaws in ADM design and implementation are identified and addressed before decisions are made that affect individuals."

"This could include ensuring that any use of ADM systems in administrative processes is consistent with the principles of administrative law," he said.

AI's potential to generate deepfakes, material that has been created or altered using technology, was also a concern.

"These can be used as a tool for sexual exploitation, abuse, and harassment online," the Assistant Minister said.

He highlighted the recent introduction of the Criminal Code Amendment (Deepfake Sexual Material) Act 2024 which strengthens existing criminal offences and creates new offences targeting the creation and non-consensual sharing of sexually explicit material online, including deepfakes.

Recalling how the use of computers in the 1980s revolutionised yacht design during the America's Cup, Gorman said, "I see that excitement again in the possibility of Artificial Intelligence. To unlock the potential of people wherever they live.

"But it is also a time to stop, and to carefully consider the potential hazards and pitfalls as we move forward."