1. The new risks
- Prompt injection and jailbreaks.
- Training-data poisoning.
- Model extraction/theft and membership inference.
- Insecure output handling and excessive agency.
2. How to test
Adversarial-ML and LLM red teaming, mapped to MITRE ATLAS, probe these behaviours across the model and its data supply chain.
3. Govern it
ISO 42001 provides a certifiable AI management system for responsible, well-governed AI.
How CyberSigma helps
Our AI/LLM security testing uses MITRE ATLAS to find prompt-injection, poisoning and model-theft risks, with a mitigation roadmap.
This guide is educational and not legal advice. Requirements evolve — validate specifics against the current standard or regulation for your situation.