AI
Securing AI Agents: Tool Allow-Lists, Scoped Credentials, Sandboxing and Audit Logs
Break a tool-calling agent with a hidden prompt injection in a local lab, then fix it with a deny-by-default tool…
AI
LLM security, prompt injection and AI-assisted tooling.
AI
AI Supply Chain Risk: Unsafe Model Files (Pickle) and Vetting Open-Source Libraries
A benign pickle payload proves that loading an untrusted model file can execute arbitrary code — then a picklescan +…
AI
Prompt Injection Explained: Build a Vulnerable LLM App, Break It, Then Defend It
Build a small local assistant, break it with direct and indirect prompt injection (including markdown-image and tool-call exfiltration), then add…
AI
Detecting AI-generated phishing emails
AI-written phishing emails no longer have the broken grammar that used to give them away. The tells have shifted from…