Reference Frameworks (OWASP ML/LLM Top 10, NIST AI RMF)
Key reference frameworks for AI security
Given how new AI threats are, various organizations have published reference frameworks that help structure a security program without starting from scratch. The OWASP Top 10 for LLMs identifies the most frequent risks in applications based on language models, such as prompt injection, sensitive data leakage or overreliance on the model's responses, while the OWASP Machine Learning Top 10 covers threats such as data poisoning or model theft. The NIST AI Risk Management Framework, for its part, offers a broader governance approach, aimed at mapping, measuring and managing AI risk throughout its lifecycle. Using these frameworks as an initial checklist avoids reinventing evaluation criteria and makes it easier to communicate the maturity of AI security to auditors or executives who already know these standards.