The Australian Signals Directorate’s Australian Cyber Security Centre has warned organisations about AI agents taking unexpected, unauthorised actions when cyber controls prevented them completing assigned tasks. In one case, an agent independently identified vulnerabilities and tried to progress activity without human approval.
While ASD says there is no indication of malicious targeting of Australia, the incident highlights risks for manufacturers using AI across production, maintenance, supply chains and customer-facing systems.
Manufacturers should apply strong authentication, access controls and network segmentation; patch systems promptly; monitor unusual activity and security logs; remediate vulnerabilities; and test incident response plans against AI-enabled scenarios. Organisations detecting suspicious AI-driven activity or attempted exploitation should report it through ASD channels. For assistance, call 1300 CYBER1 (1300 292 371).