Australia has been told to “harden its defence” against rogue or malicious use of AI in day two of parliamentary hearings into how the technology could be safely rolled out Down Under.

Global Shield Australia labelled advanced AI a “catastrophic risk” to Australia and urged the committee in Sydney that the government needed to invest in strategic AI capabilities, ramp up international engagement, and legislation guardrails and monitoring requirements.

The non-profit organisation — which has headquarters in Washington DC — gave a sobering assessment of ways malicious actors were using AI to illustrate to Australian politicians how important it was to respond correctly.

Australian Global Shield director Devon Whittle warned that already there had been evidence of foreign actors trying to use advanced AI to develop biological agents and weapons.

“Risks are increasing, whether it’s from biological weapons, whether it’s from cyber attacks, whether it’s from the misalignment problem,” he said.

“We need to be ready and prepared to respond when things go wrong, and there are things in that space Australia can be doing.”

He referenced a report by frontier AI company Anthropic a fortnight ago which detailed actions they had disrupted through their models — which included the development of the avian influenza virus.

“There were four or five examples of regimes trying to use models to support biological development practices,” Mr Whittle said.

“So, we see examples of, right now, people trying to use these models to undertake really dangerous actions.

“The problem is the broader that these capabilities are diffused through society, the more chance you’ll find that one disaffected person or that one rogue state who suddenly has the capability to put in place actions.

“We know that in the past, for example, Al Qaeda had a program to try to develop a biological weapon. The only reason that stopped is they couldn’t get the PhDs Required to put this program in place.

“Now you have a PhD in your pocket.”

Mr Whittle said while the organisation released a string of recommended actions and legislation in a report released in August, he would highlighted introducing a mandatory reporting framework and screening imported synthesised DNA products as two critical measures.

“AI models are giving new capabilities to malicious actors to develop biological agents and weapons,” he said.

“But there’s a really simple way for Australia to have one little safeguarding place, which is to screen the imports of synthesised DNA products when they are at the border.

“At the moment, that doesn’t happen.

“California has just put in place a law that would require screening of synthesised DNA.

“In Australia, this change could happen basically at the stroke of the pen of the Secretary of Department of Agriculture.”

Mr Whittle said reporting of AI agent breaches of unauthorised data, similar to the Services Australia “hack” of Medicare statistics in June, should be a “no-brainer” and welcomed the government exploring legislation on it.

He said every day, Australia companies across aviation, automobiles, pharmaceuticals, and more were forced to report when they cause harm and it should be no different for AI.

His view echoed several witnesses at the first round of hearings in Sydney on Tuesday, which included tech giants OpenAI and Anthropic who agreed that a legal reporting would provide clarity.

The committee heard from Australian’s largest union for retail, fast food and warehouse workers in their morning session, which detailed the ways AI was already impacting some of their more than 200,000 members.

The Shop, Distributive and Allied Employees’ Association (SDA) chief of staff Robert Tonkli said workers were increasingly under digital surveillance or facing automated task request or rosters which had no consideration for human needs or behaviour.

“For our members in these industries, AI isn’t a future issue. It’s already here, built into digital work systems, it sets their rosters, allocates their tasks and times of work, and more and more, it’s also deciding whether their performance managed,” he said.

“And usually, it’s brought in with little or no consultation.”

He gave an example of a check-out worker of 30 years was almost sacked when surveillance software flagged errors, which prompted an automatic first-and-final warning.

Mr Tonkli labelled AI a “work and health safety issue” and made several recommendations to parliamentarians including that the government create a Fair Work Commission expert panel on AI.

The hearings will continue in Melbourne on Thursday.