AI-enabled cyber risk and geopolitical risk are no longer separate emerging-risk issues. They are connected board-level risks that affect operational continuity, reputation and market trust, customers, regulatory exposure, supplier resilience and the defensibility of decisions made before and during an incident.
The risk is moving faster than governance
The most important insight is not that regulators and security agencies are concerned. It is that:
- The nature of the risk has changed faster than many governance models.
- Frontier AI is:
- Compressing the time between vulnerability discovery and exploitation.
- Lowering the skill needed to conduct sophisticated activity.
- Enabling more convincing social engineering, reconnaissance, exploit development and automated intrusion workflows.
At the same time, organisations are adopting AI inside their own environments through developer tools, productivity platforms, customer interfaces, security tooling and agentic workflows. There is also an increase in vibe coded applications, many of which are not appropriately hardened.
All this creates new attack surfaces and vectors: prompt injection, excessive tool permissions, non-human identities, data leakage, insecure application and integrations, AI-assisted code risk and opaque supplier dependencies.
This is why the board challenge question has changed. It is no longer enough to ask whether the organisation has a cyber program, an AI policy, or a positive security dashboard. The right question is: can the board and management demonstrate that their current state, remediation plan and incident response posture are reasonable, proportionate and legally defensible in light of today’s, and tomorrow’s, threat environment?
The board challenge gap
For many boards, the hard part is not willingness. It is technical penetration and understanding. Directors do not need to become engineers, but they do need enough understanding, and enough information, to be able to test and challenge management. A board that receives high-level comfort, vendor assurances or colour-coded reporting without evidence of control effectiveness may struggle to show effective challenge later. Boards should be able to interrogate, at a practical level, whether critical assets are known, external exposure is controlled, vulnerabilities are patched quickly enough, privileged access is place and constrained, non-human accounts are governed, backups are recoverable, logging is sufficient, supply-chain (including fourth parties) are resilient, incident roles are rehearsed and business continuity and disaster recovery plans are adequate and effective.
The same point applies to AI. Boards need a clear view of where AI is used, what data and systems it can access, what decisions and operations it influences, which vendors and model providers are relied on, how model behaviour is monitored, and what fallback processes exist if an AI system, tool or provider becomes unavailable or unsafe.
AI can also strengthen cyber defence, but defensive AI should be deployed deliberately, with governance, human oversight, auditability and clear boundaries. Adding AI tools without understanding the operating model can increase complexity rather than reduce risk.
Geopolitical risk includes ally-country action
Geopolitical risk widens the lens. The risk is not limited to war or direct conflict. It includes grey-zone activity, disinformation, foreign interference, insider threat, sanctions and counter-sanctions, conflict spillover into cyber operations, disruption to offshore service providers, and concentration risk in technology supply chains. It also includes action by allies. Export controls, national security directions or trusted-country restrictions on advanced chips, model weights or frontier models can have direct operational and security consequences. Anthropic has publicly stated that a US government export control directive required suspension of access to Fable 5 and Mythos 5 by foreign nationals, with the practical result that the models were disabled for all customers. Whether or not an organisation used those particular models, the lesson is broader: access to critical AI capability, security tooling, developer workflows, vendor support or cloud-enabled functionality can be constrained suddenly by policy decisions outside Australia.
That risk should be built into resilience planning. An organisation may face a scenario where an AI-enabled intrusion coincides with an offshore supplier outage, a sudden loss of access to a frontier model or security feature, and accelerated regulator and customer scrutiny. Testing these issues separately is useful, but not sufficient. Boards need confidence that management can make decisions under uncertainty when legal, cyber, operational, communications and customer issues collide.
What boards and management can do now
Practical steps for boards and management are clear.
- Refresh board and executive understanding of AI-enabled cyber risk and geopolitical risk, including how AI changes attacker speed, scale and targeting.
- Map critical systems, AI tools, data flows, third and fourth parties, offshore dependencies, concentration risk and legacy technology. With respect to concentration risk, refer to Addressing IT Concentration Risk report.
- Confirm the adequacy of your resourcing (both human and technological) to support the increasing speed of necessary cyber resiliency activities.
- Test the basics: patching, identity and access controls, attack surface management, backups, logging, privileged access, incident response and business continuity.
- Review technology and AI supplier contracts for audit rights, incident notification, subcontractor transparency, resilience commitments, data handling and exit support.
- Run a severe but plausible tabletop exercise that combines AI-enabled cyber compromise, supplier disruption, regulatory notifications, customer communications, sanctions escalation and operational continuity.
The regulatory and governance message is not panic. It is preparedness. But the time horizon has shifted. The Five Eyes agencies are saying months, not years, for a materially enhanced threat environment. The Australian Prudential Regulation Authority, Australian Securities and Investments Commission and the Reserve Bank of Australia are saying boards must be able to govern, challenge and evidence resilience within this context.
How independent legal challenge helps
This is where independent cyber expert legal challenge can add significant value. A purely technical review may identify gaps, but it may not answer the question the board will later face: was the organisation's approach reasonable according to applicable legal and regulatory expectations? A legal-led review, supported by independent cyber expertise and conducted under privilege where appropriate, can test whether the current state and uplift plan are defensible against the organisation's actual risk profile, critical services, threat environment and regulatory obligations.
That process helps in two ways. First, it improves legal defensibility by creating a disciplined record of the issues considered, the evidence tested, the options assessed, the trade-offs made and the actions prioritised. Secondly, it helps boards and management demystify the risk. Independent challenge can translate technical assertions into governance questions, expose hidden assumptions, distinguish material risk from noise, and help directors understand whether management's plan is credible, sequenced and adequately resourced.
The objective is not to guarantee that an incident will not occur. The objective is to be able to show, before and after an incident, that the board and management understood the risk, considered the right information and asked the right questions, challenged assumptions, prioritised action and maintained a plan that was reasonable for the organisation's size, complexity, threat exposure and critical operations.
AI-enabled cyber and geopolitical risk are now moving too quickly for passive oversight. The organisations best placed for the next incident will be those that can explain not only what they planned to do, but why that plan was reasonable, tested and capable of working under pressure.