Australia says AI is speeding cyber threats—but its incident count does not isolate AI
Australia's cyber agency responded to 1,524 incidents in 2025–26 and says AI is lowering barriers for attackers. The annual report does not state how many incidents were AI-enabled, so urgency and attribution must be kept separate.
The short version
- What happened
- ASD's cyber centre responded to 1,524 incidents in 2025–26, 20% more than the previous year, and issued more than 4,400 potential-malicious-activity notifications. Its incident classification method changed, limiting some year-on-year comparisons.
- Who it affects
- Employees and households face more convincing impersonation and less time to respond, but the safest response remains concrete: verify unusual requests through another channel, protect accounts with phishing-resistant authentication and apply security updates quickly.
- What remains unproven
- No denominator for incidents caused or materially enabled by AI
Keep in mind
No denominator for incidents caused or materially enabled by AI
Incident classification changed in 2025–26, limiting category comparisons
Cybercrime reporting is incomplete and affected by awareness and legal obligations
Average loss figures are self-reported and span all cybercrime, not AI-specific cases
Some threat judgements rely on intelligence that cannot be independently reproduced from the public report
Source analysis · 2 linked sources. Primary analysed source date: 26 October 2026.
Keep following the evidence
Join The Impact Brief founding list for selected AI reports, their effect on people and the evidence limits. Read a sample first.
Read the full report · 4 min The evidence, methods, implications and limitations
The measured rise is real; the AI share is unknown
Australia's national cyber centre says it responded to 1,524 cyber security incidents in the 2025–26 financial year, 20% more than the year before and 36% more than five years earlier. It notified organisations of potential malicious activity more than 4,400 times, including more than 2,200 individual notifications and collective notices to over 2,100 entities. ReportCyber received more than 82,000 cybercrime reports—roughly one every six minutes.
Those numbers describe the wider cyber environment. The Australian Signals Directorate's report calls AI a defining force and says it has observed increased risks and exposure, but it does not publish an AI-attributable incident count. It also says the incident-classification method was updated during 2025–26, so some categories do not align with the previous year. The 20% increase cannot therefore be read as a measured effect of AI.[1][2]
Where the agency says AI changes the attack path
The report identifies three mechanisms rather than a new species of attack. AI can help write and tailor phishing messages, analyse public or stolen data, and reduce the effort needed to identify or chain software vulnerabilities. It says most vulnerability exploitation has historically occurred within two weeks of disclosure, with some exploitation within 48 hours of mitigation advice, and expects AI-driven tools to compress that window further. That expectation is a threat assessment, not a prospective measurement of an AI effect.
AI integration can also create a new organisational attack surface. Poorly enforced boundaries, ambiguous instructions and excessive permissions may let an agent take unsafe actions or expose information. The agency's advice is correspondingly familiar: patch faster, reduce legacy exposure, log events, use passkeys or multi-factor authentication, constrain privileges, validate outputs and preserve audit trails. AI changes speed and scale; it does not make those controls obsolete.[1]
The distinction that matters
What the annual report measures—and what it infers
Operational counts are direct observations; AI attribution is an intelligence assessment without a published incident denominator.
Observed
1,524 response incidents and 4,400+ notifications
National cyber workload increased
Assessed
AI lowers effort for phishing, analysis and vulnerability work
Threat speed and scale may rise
Not reported
No AI-enabled incident count or matched comparison
The overall increase cannot be attributed to AI
People should not mistake aggregate harm for AI attribution
The report estimates an average self-reported cybercrime cost of A$33,000 for individuals and A$93,500 for businesses, but those values come from reports across all cybercrime types and are not verified losses caused by AI. Under-reporting is likely, while self-reported costs and response activity also reflect awareness, reporting obligations and the agency's own detection. They show material harm and operational load, not a controlled trend study.
For workers handling payments, credentials or customer data, the practical consequence is less time to notice a convincing request or patch a known flaw. A stronger future assessment would publish a reproducible definition of AI-enabled activity, counts by attack stage, comparison cases without AI and confidence levels for classified intelligence. Until then, the report supports faster defensive preparation, but not a claim that AI caused the overall incident rise.[1]
Continue the story
Read next
Continue exploring cybersecurity: compare the finding and its evidence limits.
What did the agent-security recreation actually prove?
A new unreviewed paper reproduced a published JFrog privilege-escalation flaw inside deliberately vulnerable containers and connects it to the 2026 OpenAI–Hugging Face incident. The recreation demonstrates an infrastructure failure mode; it cannot establish what undisclosed training data caused the agents' behaviour.
Read the next reportExplore the full source and evidence record
What this means for people
- Employees and households face more convincing impersonation and less time to respond, but the safest response remains concrete: verify unusual requests through another channel, protect accounts with phishing-resistant authentication and apply security updates quickly.
Global context
The operational data are Australian, while the threat mechanisms and much of the intelligence are international. Other countries have different reporting duties, agency visibility and exposure, so Australia's incident rates should not be treated as a global prevalence estimate.
What the evidence does not yet show
- No denominator for incidents caused or materially enabled by AI
- Incident classification changed in 2025–26, limiting category comparisons
- Cybercrime reporting is incomplete and affected by awareness and legal obligations
- Average loss figures are self-reported and span all cybercrime, not AI-specific cases
- Some threat judgements rely on intelligence that cannot be independently reproduced from the public report
What to watch next
- A published definition and count of AI-enabled incidents by attack stage
- Evidence that vulnerability exploitation windows are shortening specifically because of AI
- Independent reporting on agent permission failures and model-testing incidents
- Whether stronger authentication, patching and monitoring reduce measured harm
Living evidence record
Impact record IAI-0FBO6IX
Evidence stage
Observed
Confidence
Supported
Reporting basis
Source analysis
Independent or research support
Not yet
Record status
Monitoring
Last checked
26 October 2026
Source trail
2 direct sources across 2 source types.
People impact
Documented in this record.
Uncertainty
Limits and next checks are explicit.
Stages describe the evidence available—not whether a technology is good or bad. See the public method.
Related-source reporting disclosure
This record analyses 2 linked source records around the same underlying development. The extra records add method, date or context, but they do not by themselves constitute independent replication of every performance claim or predicted outcome.
Evidence trail
Sources used for this report
Links checked 26 October 2026
Research question, topics and regions
What Australia's national cyber-response data can and cannot show about AI's role in the 2025–26 threat environment
Topics: Cybersecurity · AI-enabled threats · Incident reporting · Vulnerability management
Regions: Australia · Oceania
This report is labelled source analysis. We summarise and analyse source material in our own words; company statements remain attributed claims until independently supported. Translated summaries preserve the meaning of the original source and link back to it. Read our editorial standards.
Reader commentary
What does this mean where you are?
Tell us what you have seen at work, in your community or in everyday life. A question is welcome too. No account is needed; we review notes before publishing them.
Explore commentary across the portal →Published reader notes
0No published reader notes yet. You can start the evidence-led discussion above.
Prefer a private correction or response? Contact the newsroom.