Topics
Articles are grouped by the job they are about. A post can appear under more than one heading when it honestly belongs in both.
- Prompt injection — How prompt injection works, and which controls still matter once a model can read untrusted text or call tools. (23 posts)
- RAG security — How to secure retrieval-augmented generation: poisoned documents, retrieval-time policy, and why a prompt firewall is not the whole control. (6 posts)
- AI red teaming — How to red team LLM applications: threat models, tool abuse, multi-turn attacks, and structured-output failures. (4 posts)
- AI agents — Attack surface, identity, least-privilege tool access, runtime policy, and sandboxes for AI agents. (26 posts)
- Incident response — Detection, phishing, ransomware, deepfakes, and incident response for teams defending production systems. (31 posts)