Commonwealth Bank says five AI agents now help its cyber team investigate threats, cutting the time to turn threat intelligence into action from four hours to five minutes. The bank has also made a free AI-assisted code-security tool available to Australian businesses, extending the story beyond its own security desk.
Watch Desk analysis
What happened
CommBank’s Agentic Shield system divides cyber investigations among specialised agents: one looks for suspicious activity, another gathers information, while others document incidents, check investigations and look for gaps in threat detection. Human cyber experts assess the evidence and decide what action to take, the bank says.
The bank also says it analyses more than 265 billion signals for potential threats each week, up 33% on a year earlier. Separately, its free CodeCheck tool guides AI coding agents through analysing software and investigating possible vulnerabilities. It currently supports OpenAI Codex and Anthropic Claude Code. Engineers must review findings and choose what to do; the tool can make mistakes, and use of the coding agents or models may cost money. CommBank has published details of the Agentic Shield deployment and CodeCheck.
Why it matters
The five-minute figure is a striking operational claim, but it is about turning intelligence into action, not proof that attacks are detected or stopped in five minutes. Even so, it gives a concrete example of AI agents being used to move work through a security operation, with people retaining responsibility for decisions.
CodeCheck takes that idea into software development: businesses can try an AI-guided vulnerability review without buying the tool itself. The catch is that it needs a supported coding agent, and its findings still need an engineer’s judgement. Free software, not free cyber expertise.
Our read
This is more than a promise to “use AI for security”: CommBank gives a specific workflow, a reported time change and a tool other businesses can download. Treat the speed improvement as the bank’s account of its own system, not a general benchmark. For teams trying CodeCheck, the sensible test is whether it finds useful issues without drowning engineers in false alarms.
What to watch
- Whether CommBank shares measured results on the quality of Agentic Shield’s investigations and actions.
- Which software projects CodeCheck helps identify vulnerabilities in, and how engineers assess its findings.
- Whether support for additional AI coding agents follows.
Discussion spark: Should AI agents be allowed to take more direct action during a cyber incident, or should they remain investigators while people make every operational decision?
Sources and evidence
- CommBank deploys five-agent AI shield as cyber threats accelerate (18 November 2026, 00:00 UTC)
- CommBank launches CodeCheck to help Australian businesses find software vulnerabilities (18 November 2026, 00:00 UTC)
- Why AI adoption alone won’t solve Australia’s productivity problem (18 November 2026, 00:00 UTC)
- A conversation between CBA's Matt Comyn and Microsoft's Satya Nadella (18 November 2026, 00:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.