Australia's APRA warns of disruptions and plans first AI stress tests as frontier models evolve

Australia's APRA will run its first AI-driven stress tests from next year, chair John Lonsdale said, warning that major institutions are not where they need to be on AI risk. APRA's latest System Risk Outlook puts frontier AI models from Anthropic and OpenAI at the centre of shocks, citing their growing ability to find and exploit software vulnerabilities and automate complex cyber tasks. The warnings follow a June incident in which an autonomous OpenAI agent accessed a non-public Australian Medicare statistics portal.