Independent senator David Pocock has blasted the federal government and renewed calls for dedicated AI safety laws after an OpenAI agent breached a Medicare statistics portal.

'This is not a new problem,' Senator Pocock told ABC’s Afternoon Briefing on Friday while criticising Australia’s lack of dedicated AI legislation. 

'For many years now, experts have been saying we need action on this. The crossbench has been pushing them. 

'The government then said, "We’ll have some guardrails, we’ll put them on a website"... that’s not enough. Clearly, it’s not. This is a huge challenge facing us.'

The Albanese government proposed mandatory guardrails for high-risk AI in 2024 and considered introducing dedicated legislation to enforce them. 

However, this proposal was shelved in December 2025, with the government’s National AI Plan instead relying largely on existing laws and voluntary guidance.

On Friday, Assistant Minister for Science, Technology and the Digital Economy Andrew Charlton told the ABC the government hoped to pass legislation mandating AI safety standards in early 2027.

Pocock claimed the government had initially shelved its earlier proposal under industry pressure.

'It looks pretty silly now,' he said. 

Pocock is a member of the Joint Select Committee on Artificial Intelligence, which was established by resolutions of the Australian House of Representatives and Senate on August 20.

The committee is due to deliver a report by November 30 on the opportunities and risks presented by AI, including its interaction with copyright and intellectual property laws and the adequacy of Australia’s existing legal and regulatory frameworks. 

Pocock said a former government expert panel on AI had recommended continuous monitoring of 'high-risk government websites and datasets'. 

'It's great to see the PM on the world stage saying we need global cooperation on this ... we absolutely do ... but it's a bit rich that at home they have shelved plans for an AI Safety Act.'

Pocock also criticised the government for encouraging large-scale investment in data centres without what he considered adequate regulation.

'We're seeing dozens and dozens of data centres being approved with basically no guardrails ... [I'm] hearing from communities who are having data centres built 20 metres from their front door. That's ridiculous.

'These companies will do whatever they can to build data centres... in Sydney the NSW state government has said these centres are state significant projects, so they skip all of the council consultation and the ability for people to ask questions about whether a proposal stacks up or not.

'We're told [AI data centres are] one of the greatest things that's going to happen to the country... once they're built, what's the long-term return for Australians? I still haven't had a good answer to that.' 

When questioned by Afternoon Briefing host Sarah Tomevska about the government’s new Intergenerational Report, which assumes average annual productivity growth of 1.2 per cent and identifies AI as a potentially significant contributor, Pocock said he was not convinced. 

'I don't think that's enough of an argument to say we're going to give up land, water, electricity, have big multinationals set up shop here ... once the big construction boom is over, what's the long-term return?'

He added that he believed companies like OpenAI should be welcome to operate in Australia 'as long as [our] laws and regulations are up to scratch... currently they're not.

'This Medicare breach highlights that. It seems very unlikely under our current laws that you could actually press any charges.

'If any Aussie did that I'm sure they would be on their way to gaol shortly, but because it's a big AI company and they say, "these are autonomous bots we've created and we didn't know or direct them to do it", that doesn't really cut it.' 

Pocock's comments come in the wake of Prime Minister Anthony Albanese's address to the United Nations in New York on Thursday, where he urged world leaders to cooperate on challenges posed by artificial intelligence and climate change.

On Saturday, AAP reported Albanese again called for an international response after OpenAI confirmed its agents had accessed US education and commerce department sites.

The company said only publicly available information had been accessed. 

'There needs to be an appropriate national response as well as an international response to make sure humans stay in charge of this new and emerging technology,' he said.

'It's up to OpenAI to state why it is then that there are now multiple cases where this has occurred.'

Albanese added he was not surprised by similar activity occurring in the US.

'If you have an AI agent act independently in the way it did here in Australia, it would make common sense that would have occurred in other places.'