AI security vulnerabilities: what it is, and what’s happened in Australia
Vulnerabilities in AI systems, software development toolchains, and hardware that can be exploited, resulting in unauthorized access, data and privacy breaches, or system manipulation causing unsafe outputs or behavior.
Definition and category 2.2 from the MIT AI Risk Repository, used under CC BY 4.0.
How experts rate it
With mitigations — 8.8%The difference between the two estimates — 6.9 points
Mean probability of a catastrophic outcome by 2030 — more than 1 million deaths, or more than USD 100 billion in financial loss, or comparable civilisation-scale intangible damage. Across 272 experts who took part. How this was measured
The study tabulates 5 of its 23 figures; this one was read from MIT's published chart on 18 August 2026.
Ranks 8 of 23 rated risks. This study published no interval for it, so the position of neighbouring risks is not meaningful. Experts put the difference at 6.9 percentage points.
Rank 8 of 23 by expert-rated severity, at 15.7%. See all 23 ranked
Evidence
- Evidence
GPT-4 successfully exploited 87% of one-day vulnerabilities when given CVE descriptions, compared to 0% for GPT-3.5 and all other tested models, demonstrating rapid capability scaling in AI-enabled cyber offense.
- Evidence
In roughly one year between model releases, AI capability for autonomous website hacking jumped from near-zero (GPT-3.5 could hack only 1 of 15 vulnerabilities) to highly capable (GPT-4 successfully hacked 11 of 15 vulnerabilities). Every open-source model tested achieved zero successful hacks.
- Evidence
The Australian Signals Directorate responded to over 1,200 cyber security incidents in 2024-25, an 11% increase. Critical infrastructure was 13% of them, up two points.
- Evidence
GPT-4 exploited 87% of tested one-day vulnerabilities when given the CVE description, and 7% without it. No other model or scanner tested managed any.
- Evidence
AI introduces novel vulnerabilities including prompt injection and poisoning attacks
- Evidence
Within the next term of government, we could face a world where simple actors can deploy thousands of highly capable AI bots working autonomously to find and exploit vulnerabilities.
What's been called for
- Not in place
Mandatory reporting of serious AI incidents
Called for by 10 organisations
Serious AI incidents and near misses should be reported to government, so regulators can see what is happening and act.
What politicians have said
Nothing here yet. Heard a politician say something about this? Add it.
What Australians say
Nothing here yet. If this matters to you, say why in a sentence.
What people ask
No questions yet. Ask one.
Think Australia should act?
Your MP and senators decide what Australia does about AI. A few lines in your own words about AI security vulnerabilities is enough to tell them it matters to you.
Write to your representative