Large Language Model (LLM)

A type of AI trained on large amounts of text to generate and process natural language. Many of the prompts in your Briefcase are aimed at large language models, and their most relevant failure mode for governance is that they can produce fluent, confident output that is wrong, which is precisely why every Briefcase card names a failure mode and every artifact carries a human check.

A deep learning model trained on enormous quantities of text (and, increasingly, other data types for multimodal models) to predict the statistically most likely next piece of content, given everything that came before it in the prompt and generated response so far. LLMs power tools like ChatGPT, Claude, and Gemini.

A deep-learning model trained on vast text data that can generate, analyze, and transform text; examples include Claude, GPT, and Gemini.

An AI system trained to generate human-like text by predicting likely continuations. The support-bot failures anchoring this topic are LLM failures; the fabrication, variance, and silent-error modes are especially characteristic of LLMs.

An AI system that generates text by predicting the next words, given the text it has been shown. Because it reads instructions and data in the same channel and tends to follow the last convincing instruction, it is vulnerable to prompt injection by design, not by defect.