The PM wants to show action on AI, even if the answers aren't there yet

The details were thin, but the vibe was clear.

Before departing this week for a round of international summits, the prime minister announced a new set of "mandatory AI safety and security protections".

In truth, these were more a set of worthy principles to somehow hold someone somewhere to account when something goes wrong — like an AI agent going rogue and hacking into a government website.

But exactly who (or what) would be held responsible, and how they would be punished, remains a work in progress.

Acting Prime Minister Richard Marles tried valiantly yesterday to explain how this might work. In theory.

"I think this will be a case of looking at the prompts that are involved in the circumstance, the way in which the AI agent is constructed … and ultimately bringing that back to the decisions, the human decisions, that have been taken by the owners of this technology or the developers of this technology," he told the ABC's Hamish Macdonald.

"But, you know, there's a lot of work which needs to be done to develop this."

Still, this was about reminding Australians, who are growing increasingly nervous about AI, that the government is on it.

The prime minister reprised the line he used at the UN General Assembly in September: "Our task is to make sure that we shape AI rather than AI shape us."

This week, there's been a growing sense the technology is the one doing the shaping.

Qantas announced plans to outsource and offshore 850 jobs as part of a plan to "unlock innovation and accelerate our use of AI", CEO Vanessa Hudson told staff in her email.

Jimmy Barnes and other artists stepped up warnings that AI firms are trying to "steal" their work.

Communities continue to speak out about their concerns over noisy and energy-hungry data centres.

Then there's the threat of AI to humanity itself.

Back in September, tech company leaders issued a series of warnings about "losing control" of AI and the risk to "humanity as a whole".

Donald Trump dismissed it as a "hoax". And, even after last week's US midterm elections backlash against data centres (among other things), there's little sign the US president has shifted his view.

In an interview with CBS's 60 Minutes, Trump made only a small concession on AI. "I don't mind guardrails," he noted, before dismissing the need for any new rules.

Trump is determined to "win" the AI race.

"It's all between us and China," he said. "There's nobody else."

Trump is unmoved by the domestic political mood in the US on AI and dismissive of the complaints from other countries grappling with the impact of this technology.

Despite this, Albanese and other middle power leaders intend to keep pressing.

The prime minister says AI safety will be a "major topic of discussion" for him at the ASEAN, APEC, and Quad leaders' meetings in the Philippines and China, where he will come face to face with Trump again.

Amid rising community concern over AI — and as both sides of politics in the US now contemplate life after Trump — it's an open question how future presidential hopefuls may approach the issue.

For Australia's part, Albanese sees the midterms result as a reminder of the need to build "social licence" for AI.

His promise to "shape AI" comes at a time when many feel they have absolutely no control.

Albanese is trying to calm fears and talk up the "potential benefits", listing economic and productivity gains and the revolution in healthcare. A vital national security imperative for access to the best AI models to defend against cyber attacks could be added to that list.

Labor knows there is no holding back the tide on a technology that's only going to become a bigger part of our lives. It just wants to avoid that tide turning into another wave of anger directed at the government, on top of the frustration voters are already feeling over the cost of living.

The prime minister wants to show he's acting, even if the answers aren't all there yet.

Who will be held responsible when AI goes rogue? The designer, or the user of the model? The tech company executive in Silicon Valley?

What penalties could realistically be imposed, and will Australia have the power to impose them?

David Speers is national political lead and host of Insiders, which airs on ABC TV at 9am on Sunday or on iview.