🧠
Language Models (LLMs)
NIST AI RMFOWASP KC1AISVS C8NIST 800-53
The core cognitive engine or 'brain' of agentic systems, responsible for understanding, reasoning, planning, and generating responses. (NIST: Data, Model; OWASP: KC1; AISVS: C8)
Subcomponents
- →Large Language Models (LLMs)Pre-trained foundation models for reasoning and generation.
- →Multimodal LLMs (MLLMs)Models that process text, images, audio, and more.
- →Small Language Models (SLMs)Smaller, focused models for specific tasks.
- →Fine-tuned ModelsModels trained for specialized domains or tasks.
Security Implications
LLMs can hallucinate, be manipulated, or leak sensitive data. Ensure prompt validation, output filtering, and model alignment.
Implementation Considerations
Apply prompt hardening, output validation, and continuous monitoring. Reference: OWASP KC1, AISVS C8.
Related Components
NIST 800-53 Rev 5 Controls
🗂️ Framework References