Technical
Large Language Model (LLM)
An AI model trained on vast text data that can understand and generate human language.
A large language model (LLM) is a type of AI model trained on billions of text documents that can understand, generate, summarize, translate, and reason about natural language. LLMs power most of the AI tools that enterprises encounter today: chatbots (ChatGPT, Claude), coding assistants (GitHub Copilot, Cursor), writing tools (Jasper, Copy.ai), and AI features embedded in SaaS platforms.
From a governance perspective, LLMs introduce unique considerations:
- They can memorize and potentially reproduce training data
- Their outputs are probabilistic, not deterministic — the same input may produce different outputs
- They can be manipulated through prompt injection
- They may produce confident-sounding but factually incorrect outputs (hallucinations)
- Different versions of the same model may behave differently
Why it matters
LLMs are the technology behind the AI tools your employees are already using. Understanding their characteristics — and limitations — is essential for writing effective AI policies and conducting meaningful vendor risk assessments.