NAB Puts Agentic AI Security to the Test in Regulated Industry

National Australia Bank (NAB) is taking a proactive approach to ensuring the security and operational safety of its agentic AI platform. The bank has been aggressively building out its AI capabilities, with a focus on designing and evaluating agentic AI systems that can operate safely in one of the most heavily regulated industries on earth.

In April 2026, NAB launched its first dedicated AI Science team, led by George Mathews, to develop the frameworks, evaluation methods, and architecture patterns necessary for deploying agentic AI agents. The bank has already seen success with OpenAI-based agents in document processing, cutting review time from 45 minutes per deed to approximately 1 minute.

NAB's collaboration with Harness suggests that compliance-as-code is becoming essential for selling into regulated industries. With a 90% adoption rate across divisions, the demand side is clear. However, the question remains whether guardrails, testing frameworks, and governance structures can keep pace with deployment that is already well underway.

Agentic AI systems are designed to take actions autonomously, processing documents, initiating workflows, and interacting with customers without human approval at each stage. A single agentic AI failure could trigger regulatory scrutiny from the Australian Prudential Regulation Authority, reputational damage, or both.