The Commonwealth Government has outlined a proposed “systems regulation” model for frontier artificial intelligence, drawing on prudential supervision, aviation safety, critical-infrastructure security and workplace health and safety.
In a speech on 8 October, Assistant Minister for Science, Technology and the Digital Economy Andrew Charlton said developers of the most capable models could be required to run rigorous processes to identify, test, report, and manage risks—and be accountable for whether those processes work. Government would set the required standard without attempting to prescribe every technical control or anticipate every hazard.
A policy direction, not enacted law. The government is developing National AI Standards legislation, and the eventual scope, enforcement powers and thresholds for identifying regulated frontier models remain unsettled.
The approach is nevertheless consequential. It would shift legal scrutiny from individual harmful outputs to the quality of the developer’s entire safety system. A company could therefore face questions about testing, monitoring, escalation and incident prevention even before a catastrophic failure occurs.
For lawyers, that points to a familiar regulatory task: establishing whether an organisation has a defensible governance system, not merely a collection of voluntary policies. Evidence of risk assessments, model evaluations, permission controls, incident logs and executive accountability may become central to compliance and litigation.
Sources: ministerial speech; ABC News.
Leave a Reply