GUEST OPINION: In security, artificial intelligence (AI) has moved well beyond hype. Organisations across Australia already use AI to improve threat detection, streamline decision-making and reduce the operational burden on security teams. The real challenge now isn’t whether to adopt AI in security, but how to implement it safely, responsibly and at scale.
AI’s potential in security is enormous but unlocking it requires closer collaboration between those developing AI-driven tools and those accountable for protecting the organisation.
The goal is not only to create powerful AI models for threat detection, but to ensure they are deployable, observable, and aligned with tangible security outcomes.
From promise to practice
AI is already reshaping both cyber and physical security. It can detect anomalies faster than humans, unearth insights from vast datasets and automate routine tasks inside security operations centres. But without the right foundations, AI can introduce new risks, from data exposure and model drift to opaque decision-making and compliance challenges.
That’s why success with AI in security depends less on tools and more on alignment. When CTOs and CSOs work in silos, AI initiatives often stall. When they work together, organisations can move from experimentation to meaningful outcomes.
The CTO lens: Building resilient AI foundations
For CTOs, AI is first and foremost an engineering problem. Security systems are mission-critical, and AI must operate reliably within them. That means designing architectures that can scale, choosing where to run the model, and ensuring performance remains consistent under pressure.
Data is central to this effort. AI models are only as good as the data they’re trained on, and in security environments that data is often sensitive. Establishing strong data governance, lineage and validation processes is essential, not just for accuracy, but for trust. A global study on trust in AI by KPMG in collaboration with the University of Melbourne found that while half of Australians (50%) use AI regularly, only 36% say they are willing to trust it, and 78% remain concerned about potential negative outcomes.
Equally important is operational discipline. Mature Machine Learning Operations (MLOps) practices help teams manage model lifecycles, control changes and maintain visibility into performance. In security contexts, where failures can have real-world consequences, observability isn’t optional.
The CSO lens: Managing new dimensions of risk
Advertisement
While CTOs focus on platforms, CSOs focus on protection. AI introduces new threat vectors that traditional security programs weren’t designed for, including adversarial attacks, data poisoning and model manipulation.
Protecting sensitive data, testing models for robustness and securing the AI supply chain are now core security responsibilities. And as always, so is privacy. When applying AI to areas like video analytics and identity systems, compliance with Australian and global regulations must be built in from the start.
In Australia, where data sovereignty, privacy and critical infrastructure protection are major concerns, organisations must embrace AI systems that are transparent, auditable and aligned with local regulatory requirements. This includes understanding where data is stored and processed, how models are trained and how decisions made by AI systems can be explained and challenged when necessary.
Crucially, CSOs also play a role in setting boundaries, determining where automation is appropriate and where human oversight remains essential.
Stronger together: shared accountability
The most effective organisations treat AI in security as a shared responsibility. CTOs and CSOs jointly define risk tiers for use cases, establish governed pathways from pilot to production and agree on controls such as change management, rollback mechanisms and approval processes.
Cross-functional governance, involving legal, privacy and operational leaders, helps ensure AI deployments align with organisational values and risk appetite, not just technical feasibility.
A pragmatic path forward
AI adoption in security isn’t a one-off. It’s an ongoing journey that rewards a measured, outcomes-driven approach. Organisations that start with clear objectives, invest in strong foundations and prioritise collaboration are far better positioned to realise AI’s benefits without compromising trust.
AI can absolutely strengthen security, but only when it’s implemented with the same rigour, accountability and partnership that underpin any critical system. For Australian organisations navigating an increasingly complex threat landscape, that collaboration may be the most important control of all.
Advertisement
https://itwire.com/guest-articles/guest-opinion/unlocking-ai-in-security-a-view-from-the-frontline