AI security & prompt-injection defense

Threat-modelling AI systems

OWASP LLM Top 10 (2025) — the canonical risk list

Prompt injection defense

Jailbreaks & adversarial robustness

Guardrails frameworks — pick by stack, not by hype

Red-teaming with PyRIT + Garak

Tool-use safety + sandboxing

Supply-chain security: model + data + plugins

Privacy & data leakage

Compliance & rollout — NIST AI RMF, EU AI Act, ISO/IEC 42001