All AI Security & AI Safety Posts
-
AI Security
AI Security 101: Where Each Attack Belongs and What Contains It
The mechanisms behind AI attacks have not changed since the spam filters of 2008. What changed is the blast radius. A map of where each attack belongs, what contains it, and what the evidence actually supports.
Read More » -
Leadership
Why We Seriously Need a Chief AI Security Officer (CAISO)
With AI’s breakneck expansion, the distinctions between ‘cybersecurity’ and ‘AI security’ are becoming increasingly pronounced. While both disciplines aim to safeguard digital assets, their focus and the challenges they address diverge in significant ways. Traditional cybersecurity is primarily about defending digital infrastructures from external threats, breaches, and unauthorized access. On the other hand, AI security has to address unique challenges posed by artificial intelligence systems,…
Read More » -
AI Security
Neural Trojans: Backdoors That Survive Safety Training
A backdoor is a specific trigger implanted in a model that stays dormant until it fires. The 2024 result that matters is that standard safety training did not reliably remove one, and adversarial training taught the model to hide it better.
Read More » -
AI Security
Model Fragmentation Is Model Sprawl, and Nobody Has the Inventory
A quantised derivative of a model you scanned is a different artefact that you did not scan. Every variant carries its own provenance, its own vulnerabilities and its own blast radius, and the first control is an inventory most organisations do not keep.
Read More » -
AI Security
Model Evasion: Why Attackers Rarely Need Adversarial ML
Evasion research and evasion practice have been pointing in different directions for years. The attacks work in the lab, the constraint that makes them hard is a defensive asset almost nobody uses, and real attackers mostly do something cheaper.
Read More » -
AI Security
AI and Canada: Pioneering Innovation, Searching for Homegrown Success
It’s easy to forget, amid the hype around Silicon Valley’s AI giants, that many of the foundational breakthroughs of modern AI were born in Canada. In fact, two of the three “godfathers of AI” - Yoshua Bengio and Geoffrey Hinton - built their careers at Canadian universities (Université de Montréal and University of Toronto, respectively). The third, Yann LeCun, did seminal work at Bell Labs…
Read More » -
AI Security
Homomorphic Encryption for ML: What the Overhead Buys
Homomorphic encryption lets an untrusted server compute on data it cannot read. The overhead is severe enough that the set of machine learning workloads where that trade is correct is small, specific and worth naming precisely.
Read More » -
AI Security
What 250 Poisoned Documents Actually Proved
Two hundred and fifty documents planted the same backdoor in models from 600M to 13B parameters. What it planted was gibberish on a trigger word. The poisoning that should worry defenders is somewhere else entirely.
Read More »
