After a former Anthropic worker claimed the technology's builders "earnestly believe that it could kill us by the end of the decade," the company's chief executive Dario Amodei, OpenAI boss Sam Altman and Elon Musk on Sunday called for a major slow-down including independent monitoring of AI models and global regulation.
But Australia could still walk the tightrope between harnessing and containing the technology, government ministers insisted at a conference hosted on Monday by the Australian Strategic Policy Institute.
"I take that at face value, but I don't think we should care very much about what individual executives say or what individual executives claim they want to happen," Assistant Digital Economy Minister Andrew Charlton told the attendees in Canberra when asked about the warnings.
"There are a wide variety of scenarios out there, and some of those scenarios are extremely dangerous."
Dr Charlton and Industry Minister Tim Ayres have touted the government's yet-to-be-legislated AI framework, announced by Prime Minister Anthony Albanese in July, in safeguarding Australians against potential dangers.
"I do take the safety challenges here seriously," Senator Ayres told the conference.
"The Australian government's approach here (is) building capability monitoring safety in the Australian context."
Also front of mind at the conference was a string of containment breaches by experimental AI models, including a swarm of OpenAI bots which ran amok and hacked into a firm called Hugging Face during a test drive in July.
"Hugging Face was a wake-up call, not only for us, but for the whole industry," OpenAI global policy vice-president Ann O'Leary told the conference.
"This weekend was really (an) articulation of what more we can do both voluntarily and what governments can do."
The tepid language was met with dismay by Greens senator David Shoebridge, who argued the government should be erecting guardrails much more aggressively.
"It is extraordinary the denial that's happening inside the Labor Party at the moment on this," he told reporters in Canberra.
"Are they not hearing the whistleblowers that have been saying this, not just in the last 12 or 24 hours, but for years?"
Other conference attendees were less circumspect on the importance of Australia getting a grip on AI to ensure governments had the same calibre of weaponry in their cyber arsenal as criminals and adversaries.
"The last 12 months for us have been an incredible roller-coaster," Australian Signals Directorate head Abigail Bradshaw said, describing the new vulnerability of systems previously impervious to years of human effort.
"That roller-coaster ride is not over."
