Commonwealth Bank: In-House AI Tackles Cyberattacks

⚡

Key Takeaways

1The Commonwealth Bank of Australia has developed an internal AI to counter the rise in cyberattacks.

2Threat signals have increased from 80 million to 400 billion per week in six years.

3The AI reduces analysis time from two days to 30 minutes, optimizing security.

💡Why it matters — French banks, often criticized for scams, could take inspiration from this innovation to enhance their defenses.

Commonwealth Bank: An In-House AI to Combat Cyberattacks

The Commonwealth Bank of Australia (CBA) recently unveiled an innovative solution at the Gartner Summit in Sydney, developing its own artificial intelligence to protect its customers against cyberattacks. In the face of a dramatic increase in threats, the bank took the initiative to create an internal system.

An Explosion of Threat Signals

Six years ago, the CBA processed around 80 million threat signals per week. Today, that number has skyrocketed to 400 billion. This increase is partly due to the growing use of artificial intelligence by cybercriminals, who repurpose malicious code by simply altering its visual appearance.

A Tailored Technological Response

Faced with this tidal wave of threats, the CBA determined that its traditional security providers could no longer keep pace. It developed a first AI agent capable of ingesting research on emerging threats, cross-referencing it with internal data, and assessing the risk for a heterogeneous environment that includes legacy systems, cloud solutions, and SaaS. Thanks to this agent, an analysis that previously took two days can now be completed in just 30 minutes. A second AI agent focuses on detecting indicators of compromise, producing automated reports that free analysts from repetitive tasks.

Collaboration and Technical Challenges

The development of these AI agents was not without challenges. The CBA did not simply integrate a language model into its alerts. Close collaboration between the cybersecurity team and internal data scientists was necessary. The first attempt failed, and it was only by involving field analysts in the design process that a functional tool emerged. This experience underscores that AI does not replace human expertise but accelerates it.

Another major challenge was revealed by the "red team" aspect. The audit reports generated by the AI lacked determinism, sometimes producing different conclusions for the same test. To address this issue, the bank had to inject fixed checkpoints into a flow that is inherently unpredictable. Any company looking to deploy AI agents will face this same challenge.

Implications for French Banks

In France, the issue of banking cybersecurity is even more pressing as financial institutions are regularly required to reimburse customers who fall victim to fraud. The example of the CBA could inspire similar solutions, enabling faster threat detection and avoiding costly reimbursements.