AI raises the stakes: what the ASD Annual Cyber Threat Report means for security, privacy and ICT contracts

On 26 October 2026, the Australian Signals Directorate (ASD) and Australian Cyber Security Centre (ACSC) published their Annual Cyber Threat Report 2025-2026.

The 2026 financial year was – to no surprise – a year of significant threat activity. There was:

• over 82,000 cybercrime reports to ReportCyber
• an overall 16% increase in the average self-reported cost of cybercrime per report for businesses
• a 30% increase in publicly reported common vulnerabilities and exposures (CVEs)
• a significant increase in ACSC intervention, responded to over 1,500 cyber security incidents (a 20% increase) and provided more than 4,400 notifications to entities of potential malicious cyber activity (an approximate 80% increase).

Cyber security is not just a technical issue – it’s also a governance, legal and operational issue. The Report highlights three themes that organisations should be considering now: the dual role of AI as both threat and defensive tool, organisational resilience, and supply chain security.

In this note, we examine key elements of these themes and what they may mean for organisations managing cyber, privacy and technology risk.

The threat of AI

The negative impacts of AI are a significant focus of the ASD in the Report. AI is being used to accelerate the speed, scale and sophistication of cyber threats:

• (Social engineering): ‘Social engineering’ in cyber security describes the various methods used, based on human psychology and behaviour, to trick people into divulging information, providing access or performing actions. It is sometimes the primary form of attack but is also often a step forming part of a longer series of actions that may be used to, for example, gain access to a company’s data or disrupt its operations. AI can be used to support social engineering by efficiently researching targets and generating convincing phishing emails and websites. This can both lower the skill barrier for this type of attack, as well as increasing the potential scale of the activities. ASIC has seen a ‘sharp rise’ in scams using AI-generated deepfake videos of celebrities and politicians
• (Security vulnerabilities): Cyberattacks often exploit CVEs to disrupt or gain access to a system or data. AI can be used for collection of data about CVEs, reconnaissance and target selection. It can also be used to analyse vulnerabilities and create exploits, reducing the time, effort and skill required for this form of attack.  The ASD notes that the movement from discovery to exploitation, which previously may have taken between days and months, can now take days or even hours.

CrowdStrike cites, as part of their 2026 Global Threat Report, an 89% increase in attacks by AI-enabled threat actors.[1] Agentic AI is enabling threat actors to scale activities, effectively increasing the workforce available to them.[2] There have also been numerous examples of AI tools becoming the source of security vulnerabilities.[3]

The benefit – and necessity – of AI

But AI is not the enemy. In this context, the ASD gives the call to action:

’Australian organisations must embrace AI to improve resilience while preparing to counter malicious uses of AI by adversaries.’

Various forms of AI and automation have been used as part of organisations’ security posture for years.

AI tooling can support security by:

• performing repetitive tasks and analysing large volumes of security data quickly
• detecting vulnerabilities and unusual behaviour
• identifying patterns and comparing data to previous patterns and data, to provide additional insights to decision-makers.

In the last week there was the highly publicised discovery of critical vulnerabilities in ‘PHP’, backend technology used in an estimated 70% of websites and applications.[4] The vulnerabilities were identified with the assistance of AI tooling.

With AI-enabled exploitation increasing the scale and speed of threats, AI and automation becomes important to enable organisations to keep pace.

As more sophisticated security tooling becomes available – in parallel with the increases in risks – regulators will expect to see more rigorous planning, processes and tooling being implemented by organisations, especially those in high risk sectors such as critical infrastructure. The increased risk posed by, and potential benefits that can be obtained from, AI tooling is directly relevant to, for example, whether an APP entity has met their obligation under Australian Privacy Principle 11.1 to ’take such steps as are reasonable in the circumstances to protect’ the personal information they hold.

Organisations implementing AI tools should carefully consider legal, procurement and governance issues, including:

• whether the tool provider will use organisational data to train its models (and if so, if this is lawful (for example, is it a permissible secondary purpose under APP 6) and appropriate)
• the location of data storage and processing
• AI guardrails, such as audit rights, review of outputs and transparency regarding automated decision-making, and
• the security practices of, and security obligations imposed on, the AI vendor.

Resilience

The ASD notes:

’The question is no longer whether an organisation will face a cyber incident, but how prepared it is to respond. Building cyber resilience means maintaining critical operations and recovering quickly when incidents occur.’

The theme of resilience aligns well the recently announced exposure draft of the Privacy Amendment (Personal Data Protection) Bill 2026 (Cth), being the next package of proposed reforms to the Privacy Act 1988 (Cth).[5] Under the proposed new sections 26WDA and 26WDB, entities must take steps as are reasonable in the circumstances to implement practices, procedures and systems relating to the entity’s functions and activities that will, among other things, enable the entity to (take reasonable steps to) prevent or reduce harm to individuals affected by a data breach.

Resilience is not merely about preventing incidents, but also preparedness. Organisations should ensure incident response plans are documented, regularly tested, aligned with notification obligations (for example, those under the Privacy Act 1988 (Cth), Prudential Standards CPS 230 and 234, and the Security of Critical Infrastructure Act 2018 (Cth) and supported by contractual arrangements with key service providers.

Customers of ICT solutions will increasingly be looking to scrutinise an ICT supplier’s business continuity and disaster recovery plans and capability, incident response readiness, backup procedures and ransomware recovery arrangements.

Proper management of cyber risks associated with supply chains is a critical issue for ensuring the resilience of an organisation.

Many incidents affect organisations through their supply chain. With organisations having increasingly interconnected supply chains, cyber risk management requires visibility beyond the organisation’s own environment. For customers of ICT services, this means an increased emphasis in contracts on:

• security obligations
• incident notification, management and remediation
• subcontracting
• audit rights, including the ability to obtain detailed information about the ICT provider’s own supply chain
• business continuity
• insurance.

Code repositories

An important and sometimes overlooked part of an organisation’s supply chain are code repositories.

Code repositories such as GitHub are critical tools used by developers to store and collaborate on software – both open source, and private repositories. Open source packages can be widely used as components across a broad range of solutions and environments. Organisations that use third party solutions may have limited knowledge of what open source packages are included within the solutions.

The ACSC notes that it has, since September 2025, been tracking the targeting of online code repositories. These attacks involve gaining access through social engineering, compromised credentials and authentication tokens, and then:

• modifying public software packages to compromise users of the packages
• publishing credentials and private code repositories
• harvest cryptographic keys and passwords.

This highlights the need for organisations to:

• maintain an accurate inventory of the software packages (including specific versions) installed across their environments – without this, an organisation may not be able to rapidly identify and deal with CVEs
• ensure security obligations in contracts apply regardless of whether the vulnerability originates in proprietary software or third party software (including open source). While open source licensing considerations often require specific carve-outs in relation to the terms of the licence (such as rights to redistribute and in relation to copyright notices), those carve-outs should not relieve suppliers from obligations to monitor, patch, remediate and manage security risks associated with third party and open source software incorporated into the solution.

Preparation for post-quantum computing

The ASD has again focussed on the issue of cryptographically relevant quantum computers (CRQCs) and the threat they pose to encryption technologies.

You can read up on our analysis of that issue from last year here.

Key takeaway

The Report demonstrates that the threat landscape is becoming faster and more complex. With AI increasing the capability of cyber adversaries, it is an increasingly important defensive tool.

Organisations should be reviewing not only their technical controls, but also their governance arrangements, supplier contracts, incident response readiness and compliance frameworks. Cyber resilience is not solely an issue for your IT security SMEs – it is a governance, legal and operational challenge and imperative.

[1] The author notes their reliance on a summary at https://go.crowdstrike.com/2026-global-threat-report.html, wishing to avoid consenting to direct marketing to obtain the report.

[2] See, for example, https://securitybrief.com.au/story/crowdstrike-finds-ai-agents-running-autonomous-cyberattacks, reporting on recent examples identified by CrowdStrike.

[3] See, for example, https://www.cyber.gov.au/blog/when-ai-agents-take-unexpected-actions.

[4] https://www.abc.net.au/news/2026-10-22/data-breach-warning-after-ai-uncovers-critical-php-flaw/107295516

[5] You can find an overview of the proposed reforms here: /insights/a-new-era-for-privacy-in-australia-implications-of-privacy-act-reform-tranche-2/

AI use disclaimer: Artificial Intelligence tools were used on a limited basis to assist in the production of this article. All AI-assisted outputs were reviewed by the authors to ensure accuracy, appropriateness, and alignment with ethical standards.