ASD has published its Annual Cyber Threat Report for 2025-26. [1]
Here’s what’s in it, read off the report rather than the coverage.
The headline numbers
ASD’s Australian Cyber Security Centre responded to more than 1,500 cyber security incidents, an increase of more than 20 per cent on last year. The precise total, from Figure 1, is 1,524. [1]
It notified entities of potentially malicious activity on their networks more than 4,400 times, almost 80 per cent more than the year before. More than 51 per cent of individual notifications came from declassified intelligence. [1]
Of the incidents it responded to, 362 were identified after one of those notifications went out. [1]
It received more than 82,000 cybercrime reports through ReportCyber, and says plainly that most cybercrime goes unreported. It answered more than 37,500 calls to the Australian Cyber Security Hotline. [1]
That 362 is the figure we’d put in front of a board first. In 362 cases last year, the organisation found out because a government agency told it.
Don’t add those numbers together
Worth saying plainly, because a lot of write-ups this week will get it wrong. The 82,000 reports, the 1,524 incidents and the 4,400 notifications measure three different things. [1]
82,000 is public and business reporting into ReportCyber. 1,524 is what the ACSC actively responded to. 4,400 is what it pushed out to organisations unprompted. One incident can sit in more than one of those, or none.
There’s a related point in the report that’s easy to miss. ASD received fewer phone calls reporting incidents this year, but the number of incidents reported went up, including unsuccessful ones. It reads that as improving awareness and reporting culture rather than as a change in the threat. [1]
What it cost
The average self-reported cost per cybercrime report for businesses rose 16 per cent to $93,500. [1]
The breakdown is the part worth reading twice.
Small business averaged $55,400, consistent with last year. Medium business rose 45 per cent to $141,300. Large business came in at $148,100, down 27 per cent. [1]

Mid-sized firms were the only group whose average loss went up.
So the largest organisations cut their average loss by more than a quarter, and mid-sized ones saw theirs rise by nearly half. Same year, same report.
Individuals averaged about $33,000 a report, also consistent with last year. Across all reports the average self-reported cost of cybercrime was $38,130, with identity fraud, online shopping fraud and online banking fraud the most frequently reported types. [1]
Large business cut its average loss by 27 per cent. Medium business saw its average loss rise 45 per cent. Same year, same report.
Fewer big incidents, but the extensive ones went up
This is the part the headlines are getting muddled, so take it carefully.
The number of higher-categorised incidents went down, with only one C2 incident recorded all year. But extensive compromises rose from 26 to 40. [1]
Denial of service tells a similar story. ASD responded to more than 80 DoS and DDoS incidents, down more than 60 per cent, which it attributes to the absence of any coordinated DDoS campaign this year. Fewer incidents, but the ones that happened hit harder: extensive compromises from DDoS activity rose from one to seven. [1]
Ransomware incidents came in at 139, consistent with last year, while extensive ransomware incidents rose from 16 to 24. [1]

Volume was flat or falling. The extensive incidents were not.
Data breaches were 98 incidents, over 6 per cent of all incidents, up 66 per cent on the previous year. Separately, the report notes 2025 was the highest year for Notifiable Data Breaches notifications since the OAIC scheme started in 2018, at 1,205. [1]
One caveat the report puts on itself, which matters for every comparison above. ASD reviewed and updated how it categorises incidents during FY2025-26 to improve accuracy, and says some incident categories will not align with last year. [1] So treat the year-on-year severity movements as indicative, not as clean like-for-like.
Who reported
Government again reported the most incidents, which the report attributes to government’s additional reporting obligations. Federal government was 32 per cent and state and local government 16 per cent. [1]

Manufacturing entered the top 10 with the largest increase of any sector. Transport dropped out.
Then health care and social assistance on 7 per cent, professional, scientific and technical services on 6 per cent, financial and insurance services on 6 per cent, information media and telecommunications on 5 per cent, education and training on 4 per cent, manufacturing on 4 per cent, and construction and retail trade on 3 per cent each. [1]
Two movements in that list. Health care and social assistance rose to be the most frequently reporting non-government sector. Manufacturing entered the top ten with the largest increase in incident reports of any sector, and transport dropped out. [1]
Critical infrastructure
Critical infrastructure incidents were 10 per cent of all incidents, similar to last year. ASD responded to 154 of them, a decrease of almost 7 per cent. [1]
The top three sectors for critical infrastructure incidents were financial and insurance services at 30 per cent, public administration and safety at 21 per cent, and electricity, gas, water and waste services at 17 per cent. [1]

Two of those three are the sectors BayRisk works in.
Financial services being the largest slice is worth pausing on, because it’s easy to misread. Financial and insurance services is 6 per cent of all incidents and 30 per cent of critical infrastructure incidents. Different denominators, same report. If you’re in that sector, the second number is the one your SOCI obligations are measured against.
The top three initial access techniques into critical infrastructure were phishing at 36 per cent, valid accounts at 16 per cent and drive-by compromise at 10 per cent. [1]
So more than half of critical infrastructure intrusions started with either a phishing email or a working set of credentials.
The report is direct about state-sponsored activity too. It says state-sponsored actors routinely target Australia’s critical infrastructure networks, possibly for espionage, or to pre-position for disruptive and destructive effects against operational technology in the event of crisis or conflict. [1]
A coordination mechanism for financial services that nobody has written up
This one didn’t make the coverage and it should have.
The report describes CART, the cyber-attack response protocol, which coordinates the response to significant cyber incidents affecting Australia’s financial system. It brings together APRA, ASIC, the Reserve Bank of Australia and Treasury to share information, maintain situational awareness and support a common operating picture. [1]
During significant incidents, CART connects financial sector response arrangements with the broader government cyber incident management frameworks, including those led by ASD’s ACSC and the Department of Home Affairs, so that regulatory, supervisory and operational responses are aligned. [1]
If you’re an APRA-regulated entity, that’s worth knowing before you need it. In a serious incident your prudential regulator, your conduct regulator, the central bank and Treasury are already in a room together comparing notes. Your incident communications should assume that.
How they got in, across all incidents
The report maps incidents against MITRE ATT&CK. The top techniques were phishing at 39 per cent, compromised accounts at 21 per cent, user execution at 17 per cent, valid accounts at 15 per cent, then social engineering, gathering victim identity information and exploiting public-facing applications at 13 per cent each. [1]
Those overlap, because one incident can involve several.
ASD reads the frequency of user execution and social engineering as showing that humans remain an enduring target at every incident category. [1]
On vulnerabilities, it recorded successful exploitation of a vulnerability identified in 2008, and makes the point that patching legacy vulnerabilities matters as much as keeping up with new ones. [1]
One worked example of what proactive notification buys. A collective notification to 340 entities in May 2026 about the cPanel vulnerabilities resulted in more than 60 confirmed incidents. [1]
AI amplified what was already there
The report’s position on AI is more precise than the headlines suggest, and the precision is the useful part.
“AI capabilities have not introduced new cyber threats, but they have amplified and sped up existing ones, meaning action is now more critical than ever.” [1]
AI is reducing the expertise and effort required to find and exploit vulnerabilities, which the report says is likely accelerating time to exploit and increasing the number and diversity of actors able to attack. [1]
Deputy Prime Minister and Defence Minister Richard Marles wrote the foreword. “AI has been the defining force in the cyber threat landscape over the past year, allowing malicious actors to find vulnerabilities faster and exploit them at unprecedented scale, speed and sophistication. In recent weeks, new threats have emerged from the unintended actions of AI agents and model testing.” [1]
He also makes a point about convergence that’s easy to skip past: “The line separating state-sponsored espionage from sabotage and cybercrime is becoming increasingly blurred.” [1]
The five actions, and the mindset underneath them
This is the part to take to an ExCo meeting.
ASD is urging Australian industry to adopt what it calls the ‘assumed breach’ mindset, which it describes as a premeditated planning approach for when, not if, a cyber compromise occurs. Five critical actions form the focus of Cyber Action Year, the coordinated effort ASD has run with government, industry and critical infrastructure since the beginning of 2026. [1]
- Replace or mitigate legacy technologies.
- Prepare for post-quantum cryptography.
- Implement best practice event logging.
- Prepare for AI-enabled cyber threats.
- Supply chain management, including choosing secure by design.
On the second one, the report is specific about timing. Ceasing the use of traditional asymmetric cryptography by the end of 2030, or sooner, will leave an organisation at lower risk should AI speed up developments in quantum computing. [1]
These aren’t new asks. The report says ASD spent the year amplifying priorities set out in last year’s report, and that they’ll remain a key focus through FY2026-27. [1]
Two other things in the report are worth noting if you’re working through the Essentials. It confirms ASD released the Essentials series in October 2026, describing it as evolving the Essential Eight into a modern, principles-based framework tailored to different technology environments. And for critical infrastructure operators it points to CI Fortify, its guidance on strengthening resilience and preparing for incidents and disruptions to essential services. [1]
Where BayRisk lands on this
Three things here are worth more of your attention than the headline count.
The first is those 362 incidents found only after ASD made contact. Detection is still the weakest link, and for a lot of organisations the control that worked last year was a notification from Canberra.
The second is the medium business cost figure. Up 45 per cent, in a year when the largest organisations brought theirs down 27 per cent. Big firms spent money and it worked. Mid-sized firms are carrying the same obligations without the same capability, and the gap now has a number on it.
The third is the ‘assumed breach’ framing. That’s not a technical instruction. It’s a governance one, and it’s quite a thing for the national cyber agency to put in writing. If you plan for when rather than if, then detection, logging, supplier visibility and your recovery plan stop being compliance line items and start being the things that decide how bad your worst week is.
None of that needs a new framework. It needs someone senior keeping detection, third-party exposure and board reporting current all year, rather than once before a review. That’s what we mean by RiskOps, and for a lot of mid-sized organisations it’s a layer they don’t need or cant afford full time.
If you want to work through what these numbers mean for your own position, that’s a short conversation: hello@bayrisk.com.au.
References
- Australian Signals Directorate, Australian Cyber Security Centre, Annual Cyber Threat Report 2025-26, Commonwealth of Australia, 2026. Figures and quotes above are drawn from the Foreword, the Year in Review, Figures 1, 2 and 8, the Cybercrime, Ransomware and Critical Infrastructure sections, Highlight 1 (OAIC Notifiable Data Breaches), Highlight 4 (CART) and the Resilience section (Cyber Action Year). https://www.cyber.gov.au/publication/annual-cyber-threat-report-2025-26