Notes from the field.
Attack walkthroughs, research deep-dives, and lessons from securing AI in production.
Security in AI: An Introduction
How to secure AI Systems. The opening primer in our series. An introduction to how AI models are compromised, a map of the attack surface, and what you can do about it.
What we've been writing.
Misalignment
What happens when an AI gets very good at the wrong thing. A primer on misalignment, reward hacking, and interpretability.
Prompt Injection
How malicious text hijacks AI behavior. A primer on direct and indirect prompt injection, and mitigations.
Data Poisoning
The model learns from what it's fed. Poison the data, poison the model. A primer on data poisoning and mitigations.
AI Threat Detection - Runtime Defense for Enterprise AI
AI threat detection is the continuous identification of malicious behaviour targeting AI systems. Learn what it covers and how to deploy it in regulated environments.
How to Secure LLMs - a 6-step practical guide
A practical six-step blueprint for teams shipping large language models into regulated environments. No theory, no checklists for their own sake.