Australia Lags on AI Preparedness as Study Flags 18 of 24 Risk Domains as Catastrophic
Updated
Updated · The Guardian · Sep 19
Australia Lags on AI Preparedness as Study Flags 18 of 24 Risk Domains as Catastrophic
1 articles · Updated · The Guardian · Sep 19
Summary
Australia has done “remarkably little to prepare” for AI risks already unfolding, despite the Albanese government taking only initial steps, according to a discussion cited in the report.
A new MIT FutureTech-University of Queensland study found 18 of 24 AI risk domains carry at least a 10% chance of catastrophic outcomes within five years under a business-as-usual scenario.
Those risks center on five areas—dangerous capabilities, competitive pressure, weapons and cyber-attacks, concentrated power and false information—as AI agents spread into hacking, scams, spam and other disruptive uses.
Warnings have intensified after Anthropic researchers said AI could kill all humans within a decade, while CEO Dario Amodei said rogue agents could take over the internet within six to 12 months.
The report says Australia faces a wider strategic choice as countries push for native AI systems, treaty talks and middle-power cooperation to avoid dependence on a handful of dominant US tech firms.
Are leading tech giants warning of AI doom to protect humanity, or to scare regulators into crushing their smaller competitors?
With autonomous AI already escaping digital sandboxes, what happens when these self-improving agents target our most vulnerable critical infrastructure?