Ransomware activity across Australasia has jumped more than 600% and AI could push down the cost of an attack, yet New Zealand’s cyber insurance market remains soft, with ample capacity on offer.
Ensign InfoSecurity found organised crime behind 52% of observed threat activity, against 36% for state-sponsored groups and 12% for hacktivists.
Ransom remained the most common outcome, and manufacturing and industrial firms, the public sector, business and professional services, and healthcare were frequent targets.
Aon rated New Zealand’s cyber market soft in the first quarter of 2026, with abundant capacity and wider coverage available.
Threat data runs the other way. The National Cyber Security Centre (NCSC) handled 369 incidents of potential national significance in 2025/26, with 162 linked to criminal or financially motivated actors, an 18% rise.
Read more: NZ cyber threat rises as insurance market moves the other way
“Stronger cyber maturity does not remove the incentive to attack. The region has mature, highly digitalised and well-insured organisations, connected through extensive outsourcing and cloud infrastructure. That makes sectors like health, manufacturing and professional services particularly valuable to threat actors,” said Charles Spencer, general manager for Australia and New Zealand at Vectra Corporation, through which Ensign operates in Australia.
Adelphi Insurance Brokers found New Zealand ransomware claims dipped slightly in frequency in 2025 but rose in severity and cost, with business interruption the main driver, and Commvault found 34% of ransomware-hit organisations in Australia and New Zealand paid, and 36% of those payments failed.
Prices for stolen identity packages, or Fullz, have also climbed. Australia recorded Asia-Pacific’s highest observed price, and the highest observed Australasian listing rose from US$60 in 2023 to US$280 in 2026, about 367% more.
Ensign said the rise suggests criminals place greater value on identity records and related data tied to Australian organisations and individuals, and linked it to a digitally mature, relatively wealthy economy.
Read next: Ransom paid, data still lost – and brokers are asking the wrong question
Attackers are turning to indirect routes while companies tighten perimeters, the report found. Sales of initial access rose fivefold, and supply-chain compromise and insider access became more important entry points.
Suppliers, contractors and insiders offer a trusted way into better-defended firms, a particular exposure for manufacturers, many of which operate through extended supplier networks, outsourced services and shared digital platforms.
“Attacks do not always need to breach a well-defended organisation directly. A supplier, contractor or compromised insider can provide a trusted pathway into multiple organisations,” Spencer said.
Threat actors increasingly combine financial, intelligence and disruptive motives, he said, so “organisations need to understand where they sit within a wider ecosystem, not just how secure their own perimeters appear.”
In May 2026, a supply-chain attack on Canvas, a US-owned, cloud-hosted learning platform, hit more than 30 New Zealand tertiary institutions and schools. Institutions lost access to essential services for several days.
The NCSC attributes 23% of nationally significant incidents to suspected state actors. Since March 2023, Lloyd’s has required standalone cyber policies under its CY and CZ codes to exclude losses from certain state-backed attacks, though the NCSC notes state and criminal groups share tools and techniques.
To gauge frontier AI’s effect on attack costs, Ensign reviewed more than 150 models and tested 10 generally available ones over 160 runs across eight stages of a simulated enterprise attack, inside an isolated cyber range resembling a university network with widely used security controls.
The chain ran from breaching a portal and compromising a database through to establishing persistence. Several leading models completed multiple stages; all could steal credentials and move laterally to some degree, but none could evade detection tools with confidence.
Z.AI’s GLM-5.2, an Eastern model, delivered offensive performance comparable to GPT-5.6 Sol at about one-fifth of the operating cost, and open-source Eastern models produced more offensive output per dollar than Western counterparts.
Because performance is converging, Ensign said price may matter more to attackers than technical differences, which could lower barriers to cybercrime by making sophisticated tools cheaper and easier to use at scale.
“Frontier AI is fundamentally changing the speed, scale and economics of cyberattacks. If it can automate reconnaissance, identify vulnerabilities and help threat actors repeat those steps at a lower cost, it gives attackers more opportunities to find a way in,” Spencer said.
He added: “With frontier AI capabilities advancing on a roughly two-month cycle, security controls will need to evolve just as quickly.”
Browne Jacobson’s Tim Johnson notes that many policies define a hacker as a person, so some wordings may not pick up an AI attacker, and Assured senior broker Casper Rogers says insurers may revisit language similar to Chubb’s former Widespread Vulnerability Exclusion to limit aggregated exposure.
“It might be that AI just makes everybody a super cyber criminal, and that turbocharges everything,” said Jeffrey Gonlin, chief underwriter at Emergence Insurance.
