Particle.news

Frontier AI Drives Sharp Rise in Cyberattacks on Australia

Authorities warn faster, AI-powered vulnerability discovery raises the prospect of service outages that could threaten customer savings.

Overview

  • The Australian Signals Directorate’s annual Cyber Threat Report, released Monday, recorded about 4,400 suspicious-activity notices and more than 1,500 incident responses for 2025–26, an 80 percent jump in notifications year on year.
  • Advanced ‘frontier’ models can find and chain software flaws far faster than before, and agencies link this capability to recent incidents such as a global PHP outage and reports of AI agents accessing live Australian data.
  • The ASD named China, Russia and North Korea and proxy groups as persistent state-sponsored threats that routinely probe networks to steal data or preposition for disruption that could incapacitate services.
  • Officials are urging firms to prioritise security over availability, update legacy systems, rotate cloud credentials, test frontier models with Five Eyes partners and deploy AI defensively to detect and block attacks.
  • Regulators warn the human cost could be severe: APRA officials say AI-amplified attacks could put individual superannuation and bank accounts at risk while businesses face rising remediation costs and more frequent outages.