Federal government accused of AI policy retreat as US tech giants plan Australian investments

A former Labor minister says the federal government "blinked" on AI regulation, shelving plans to make it safer for consumers rather than provoke US President Donald Trump.

When he was minister for industry and science, Ed Husic was planning to bring in "mandatory guardrails" on high-risk AI as part of a standalone act that aimed to protect the community against the technology's potential harms.

After the election in May 2025, he was dumped from the ministry in a factional deal. By the end of the year his policy proposal was also gone.

"We were already attracting heat from him [Trump] and Elon Musk on social media laws. You saw the treatment of our own eSafety Commissioner by that administration and prominent figures in the US, and so we just figured we'd duck that fight."

Minister for Industry and Science Tim Ayres disputes Mr Husic's interpretation of the decision, telling Four Corners the policy shift had nothing to do with the Trump administration.

"The Albanese government has demonstrated that we are absolutely prepared to regulate big tech where it's in the national interest. We've moved with world-leading legislation to limit the access of under-16s to social media," he said.

"We've delivered an Artificial Intelligence Safety Institute. It will work across government. It will work with community institutions and with experts here in Australia."

But as big tech begins investing in Australia, there are fears the government's regulatory retreat could have serious consequences.

'A hands-off approach'

There are growing concerns about AI safety across the world, including the US, where there have been a series of lawsuits against AI giants by parents who allege their children were coached by chatbots to self-harm or suicide.

OpenAI's ChatGPT has been the subject of a study called Fake Friend, run by the Center for Countering Digital Hate.

"We created fake profiles as young people. We then went and asked it a series of structured questions, with a child who is thinking about killing themselves. Within two minutes we were able to get ChatGPT to tell us how we can cut ourselves safely," CEO Imran Ahmed said.

"Within 65 minutes it was giving us a full personalised suicide plan and drafting suicide notes for our family and friends.

A spokesperson for OpenAI told Four Corners that the wellbeing of teenagers was a "top priority".

"We have safeguards in place today … and we're continuing to strengthen them," they said.

There were a range of safety concerns at the heart of Mr Husic's plan for an AI Act, including minimising harm to children, protecting physical and mental wellbeing and preserving civil liberties and judicial rights.

However, Senator Ayres believes the existing legislation provides enough protection when properly enforced.

After abandoning Mr Husic's proposed guardrails, the federal government unveiled a National AI Plan in December last year.

The plan stresses that existing laws are the best way of dealing with AI-related risks. As an example, the plan points to what it calls Australia's "robust privacy protections".

However, Professor Ed Santow from the Human Technology Institute, who ran an inquiry into the human implications of AI while human rights commissioner, said the current privacy laws were anything but robust.

"Our privacy law is dangerously out of date. It was largely drafted before the internet, let alone the age of artificial intelligence," he said.

"If you can update our privacy law, that's probably the single biggest step you could take to protect Australians from their personal information being used against them in the age of AI."

Mr Husic is deeply concerned by the government's sudden pivot on its approach to AI, and fears it could be detrimental.

"We've done a 180-degree turn. We've gone from recognising the risk and recognising that we should do something now to having a hands-off approach," he said.

Big tech comes to town

In April, two of the world's biggest tech titans visited Australia, promising investment opportunities right when the federal government needed them most: while the economy was flat and productivity continued to lag.

Dario Amodei, the CEO of Anthropic, visited Parliament House in Canberra and signed a Memorandum of Understanding (MOU) with the government to "work together to capture the full benefits of AI for Australians".

Three weeks later, Microsoft CEO Satya Nadella met with Prime Minister Anthony Albanese and announced what he called "the largest ever company investment in Australia" of $25 billion.

Senator Ayres told Four Corners the companies wanted to expand their AI infrastructure here in a country that was safe, democratic and with plenty of open space.

"They know that Australia is a very good place for data centre investment," he said.

"We have enormous competitive advantages that a lot of economies around the world don't have, a future competitive advantage in clean green energy."

The AI companies are desperately seeking new territory to build the data centres they need to provide the computational power required to train and run their AI models. With the backlash building against data centres among US communities, big tech is looking to Australia.

"There are four things that tech companies are looking for when they're looking to build data centres," Professor Santow said.

"They need access to huge amounts of energy, huge amounts of water. It needs to be a safe, stable place, and it needs to be a country that is not subject to the restrictions that the US government is putting on the sale of chips — these incredibly powerful chips that enable data centres to do their extraordinary work."

Helen Toner, a former OpenAI board member, who is now the executive director of Georgetown University's Center for Security and Emerging Technology, believes investing in data centres could provide strategic benefits for Australia.

"In the 21st century, computing power is going to be a major source of national power," she said.

But there are significant obstacles that stand in the way of Australia becoming a potential data centre superpower.

Power play

Data centres generate extreme heat and need deep reserves of water to cool them and require enormous amounts of electricity.

Combined, the infrastructure is likely to put immense pressure on Australia's water and energy supplies and its emissions reductions targets.

Four Corners has learnt just how substantial that pressure could be.

In its MOU with the government, Anthropic stated that it "recognises the importance of expanding Australia's energy supply and transmission with a focus on firmed renewables". However, that agreement is not binding.

Sources close to the industry have told Four Corners on the condition of anonymity that senior Anthropic representatives disclosed to them the company had bold ambitions to invest in data centres that would use a large proportion of Australia's electricity output.

They said Anthropic disclosed it wanted 5 gigawatts of new compute to help train its AI models by 2030, with a longer-term goal of 20 gigawatts. That would significantly increase Australia's annual electricity generation.

Anthropic would not comment on those figures but said in a statement to Four Corners it "wants to be an accelerator and a partner in Australia's energy transition".

"We plan to scale our demand, including through new generation; cover our own costs; contribute to grid resiliency through load flexibility; and partner with the community to ensure the benefits are broad and shared," a spokesperson said.

The federal government said "discussions are commercial in confidence".

Former chief scientist Alan Finkel, who previously led the National Electricity Market Review, said longer-term goals such as 20 gigawatts could prove challenging given solar, gas and wind were the government's preferred options for expanding electricity generation.

"I'm not necessarily against [data centre investment]. This is part of the economic growth and the trajectory we're choosing, but it would be huge," he said.

"To do that in a decade would be hard. To do that in five years would be extremely difficult."

'They don't want anything to stand in their way'

Companies such as Anthropic also face another hurdle if they want to train AI models in Australia: whether the nation's copyright laws need amending.

It has already sparked heated debate, with the federal government rejecting a proposal by the Tech Council of Australia for a copyright exemption for companies to train AI models on local data last year.

But the chair of the Tech Council of Australia, Atlassian co-founder Scott Farquhar, is still vigorously pushing for changes to copyright laws.

"If I train in Australia, I need to cut a deal with every single recording artist in the entire world because of the way our copyright laws work. So without some government change … it is impossible," he told the Australian Financial Review's AI summit last week.

Dean Ormston, the CEO of APRA AMCOS,  the body that collects royalties for songwriters and composers, has accused those lobbying for such changes of endorsing the theft of intellectual property.

"I think with the Tech Council of Australia and the Business Council of Australia lobbying the government to weaken copyright rules and frameworks, they can't step away from the outcome of that, which is theft: the theft of people's creative work," he said.

Senator Ayres insists that no matter the plans of the AI companies, the government is not budging in protecting the rights of artists.

"We have a sense of real patriotism and attachment to Australian culture, and copyright protections are an important part of building that and securing that," he said.

But to Mr Husic, this push from the tech lobby is part of the reason why the government needs to take a tougher approach to AI regulation.

"Big AI wants to get even bigger and they don't want anything to stand in their way to make that happen," he said.

"If data laws, privacy laws, copyright laws, even taxation laws get in the way, they want to be able to charge through that.

Watch Four Corners' full investigation, The AI Race, tonight from 8:30pm on ABC TV and ABC iview.