Cyber risk is undergoing a structural shift. What was once episodic is now continuous, driven by rapid advances in frontier AI capability and reinforced by heightened regulatory attention in Australia. Capability disclosures from frontier AI laboratories, coupled with signals from APRA and the Australian Signals Directorate, point to the same conclusion: longstanding assumptions about cyber risk no longer hold.
Frontier AI models have already demonstrated the ability to autonomously discover and exploit vulnerabilities at a speed and scale previously limited to highly resourced actors. As these capabilities evolve, the window between vulnerability discovery and exploitation continues to compress, challenging existing risk, control and assurance frameworks.
This shift elevates cyber risk from a technical issue to a board and senior management priority. And it requires organisations to reassess risk posture, governance and operational response in line with a faster, more complex threat environment.
Why this matters for Australian organisations
Australia occupies a distinct position in this shift. As a Five Eyes partner with deep integration into allied intelligence sharing, the country benefits from early threat visibility. It also sits within the targeting envelope of state sponsored actors with access to frontier AI capabilities.
For Australian organisations, this is not a theoretical shift. It is a change in operating conditions.
What changes for the enterprise?
The response to this shift requires action across three different layers of the organisation, and none can be addressed in isolation.
Cyber risks in the age of frontier AI can no longer sit primarily with technology and cyber teams. This is a board and senior management leadership matter, and they mandate the organisation’s risk posture. They should be changing the questions they ask of management, just as audit committees should reshape reporting, anchored in operational indicators rather than control assurance alone.
APRA’s April 2026 letter reinforces this directly: boards must ‘maintain sufficient understanding and literacy with respect to AI in order to set strategic direction and provide effective challenge and oversight.’
For critical infrastructure operators, the Security of Critical Infrastructure Act imposes equivalent obligations through risk management programs. The structural shift in AI-enabled offensive capability makes these obligations materially more demanding across all sectors.
Existing technology and operational risk assessments were calibrated to a threat economy where exploit development was expensive, slow and concentrated in skilled human attackers. That assumption no longer holds in the current landscape. Risk taxonomies, materiality thresholds, third party assessments and CRO metrics all require re baselining to reflect AI compressed timelines and chained vulnerabilities.
For prudentially regulated entities, this includes reassessing the adequacy of information security capability relative to the new threat environment. For Critical Infrastructure operators, risk management programs must now account for AI-enabled threat scenarios. The principle applies broadly: any organisation whose risk settings were calibrated before frontier AI must revisit them.
The window between vulnerability disclosure and active exploitation which was once measured in months, is now measured in days, hours and minutes. Prioritising vulnerabilities by severity score alone and reviewing your external attack surface quarterly are no longer sufficient. Internal Audit must now test what was not previously tested and the assurance evidence flowing to the board must keep pace with the new operating tempo.
The Essential Eight maturity model is the baseline for Commonwealth entities and widely used across the private sector, but target maturity levels and patch timelines now require reassessment as they may no longer reflect the reality of AI-compressed exploit cycles.
Australian sector exposure
The impact of this shift is not uniform across sectors. In Australia, several industries carry disproportionate exposure.
How KPMG can help you work through the shift
KPMG Australia offers private briefings for boards and senior management, targeted reassessment of risk and control frameworks, operating model uplift across all three lines of defence, and technical readiness assessments. All services are calibrated to the Australian regulatory environment and the specific demands of this new operating regime.
Contact the team
- mark
- aaron
- greg
- ian
- anna
- charlie
Frequently asked questions
Frontier AI models have demonstrated the autonomous ability to discover and exploit zero-day vulnerabilities across every major operating system and web browser. This capability was previously the preserve of well-resourced nation-state teams. It will commoditise within 12 to 18 months. The speed, scale and sophistication of cyber-offensive capability are no longer bounded by the supply of skilled human attackers.
Boards must treat this as a governance matter, not a technology matter. The actions to take include:
- Seek a briefing on the structural shift and its implications for your organisation’s specific risk profile.
- Mandate a reassessment of your risk and control frameworks against AI-compressed attacker timelines.
- Challenge management on patch velocity, detection capability, and legacy estate exposure.
The window between vulnerability disclosure and active exploitation has compressed from months to days or hours. A meaningful share of exploitation now precedes the availability of any patch. The 30- to 90-day patching tolerances most Australian organisations operate against are no longer defensible, with critical patch response times now measured in hours, not weeks.
CPS 234 requires information security capability commensurate with the size and extent of threats. The structural shift in frontier AI means the threat environment has materially changed with existing capability assessments now requiring revalidation.
Under the SOCI Act, responsible entities must maintain critical infrastructure risk management programs that account for AI-enabled threat scenarios. APRA, the ASD and Home Affairs are expected to provide further supervisory guidance.
The Essential Eight remains the right framework. However, target maturity levels and implementation timelines require reassessment. Patch management at Maturity Level 3 (within two weeks for critical vulnerabilities) may no longer reflect the reality of AI-compressed exploit cycles. Organisations should assess whether current maturity targets are adequate given the new threat tempo.
