Australia's Commonwealth Bank built its own agentic AI threat hunting tools after finding that commercial vendors couldn't keep pace with emerging AI-powered threats. The bank's weekly threat signal volume grew from 80 million to 400 billion, making traditional defenses unworkable. Two custom agents were developed: one that ingests threat research and produces risk assessments in 30 minutes (down from two days), and another that identifies indicators of compromise. The project required close collaboration between frontline security analysts and data scientists. A key challenge was making AI outputs deterministic enough for repeatable security predictions, particularly in red team assessments. The bank's security chief warns all organizations to prepare for AI-amplified attack volumes.
Sort: