All AI Security & AI Safety Posts
-
AI Security
“Magical” Emergent Behaviours in AI: A Security Perspective
Emergent behaviours in AI have left both researchers and practitioners scratching their heads. These are the unexpected quirks and functionalities that pop up in complex AI systems, not because they were explicitly trained to exhibit them, but due to the intricate interplay of the system's complexity, the sheer volume of data it sifts through, and its interactions with other systems or variables. It's like giving…
Read More » -
AI Security
Dynamic Data Masking Protects the View, Not the Model
Masking rewrites what a query returns and leaves the stored data alone. Whoever builds your training set either reads through the mask or under it, and neither outcome gives the resulting model any privacy property at all.
Read More » -
AI Security
Every Thumbs-Up Can Become a Training Label
Flipping 0.3% of preference labels reached a near-perfect attack success rate against safety alignment. The old defence advice assumes a labelled dataset you can audit, and that is not where labels live now.
Read More » -
AI Security
Testing Can Find Backdoors. It Cannot Prove They Are Not There.
Backdoor attacks in the context of Machine Learning (ML) refer to the deliberate manipulation of a model's training data or its algorithmic logic to implant a hidden vulnerability, often referred to as a "trigger." Unlike typical vulnerabilities that are discovered post-deployment, backdoor attacks are often premeditated and planted during the model's development phase. Once deployed, the compromised ML model appears to function normally for standard…
Read More » -
AI Security
Your Guardrail Is a Text Classifier
The 2018 literature on fooling text classifiers reads like history until you notice that prompt injection detectors are text classifiers, and the same tools break them at up to 100% evasion.
Read More » -
AI Security
The Instruction Comes In Through the Camera
In 2022 a multimodal attack meant fooling a fusion model into misclassifying. In 2026 it usually means an instruction arriving through a channel your text filters cannot read.
Read More » -
AI Security
Query Attacks Buy Reconnaissance
NIST lists query access as an attacker capability, and three different attacks reach a model through it. Where the model makes a security decision, extraction is not the loss. It is reconnaissance for the evasion that follows.
Read More » -
AI Security
Differential Privacy Protects the Training Set, Not the System
What epsilon actually guarantees, why deployed values sit near 10, and what differential privacy leaves uncovered in a 2026 AI system.
Read More »
