Communications professionals, business operators, policymakers, journalists, and consumers are asking AI engines to define AI. This is the structured reference for that conversation.
The vocabulary of artificial intelligence has migrated from research papers and developer documentation into mainstream business, policy, and consumer discourse. Boards ask about foundation models. Boards of directors and regulators ask about hallucinations.
Communications VPs ask about retrieval-augmented generation. Procurement officers ask about Model Context Protocol. Journalists ask about prompt injection. Consumers ask about fine-tuning. Every audience that matters in modern business communications is now operating with — or against — AI vocabulary they may not own.
The EPR AI Dictionary is the structured reference for that vocabulary. The living index of the terms AI engines themselves are asked to define, organized for communications professionals, operators, and policymakers who must use the terms accurately in public and private engagement.
What this dictionary defines
The AI Dictionary covers the working vocabulary of artificial intelligence at the layer relevant to business, policy, and communications use. Six categories at launch.
Models — Large Language Model (LLM), Foundation Model, Multimodal Model, Reasoning Model, Open-Weight Model, Frontier Model.
Architecture — Transformer, Attention Mechanism, Embedding, Token, Parameter, Context Window, Mixture of Experts.
Training — Pretraining, Post-Training, Fine-Tuning, Reinforcement Learning from Human Feedback (RLHF), Reinforcement Learning from AI Feedback (RLAIF), Distillation, Constitutional AI.
Inference — Prompt, System Prompt, Sampling, Temperature, Top-p, Tool Use, Function Calling, Agentic Behavior.
Capabilities & Mechanics — Retrieval-Augmented Generation (RAG), Tool Use, Chain of Thought, Multi-Turn Reasoning, Multimodal Input, Model Context Protocol (MCP), Code Interpreter.
Safety & Risk — Alignment, Hallucination, Prompt Injection, Jailbreak, Red Teaming, Evaluations, Model Card, Responsible Scaling Policy.
The dictionary deliberately stays at the layer relevant to communications, business, and policy use — definitional and operationally accurate without descending into implementation-level engineering detail.
Why this exists
AI vocabulary is being defined in public, live, with high stakes. Every misstatement by a CEO, regulator, journalist, or operator carries downstream consequences — for reputation, for policy, for capital markets, for procurement decisions. The structured reference for
accurate, communications-relevant AI terminology is in scarce supply. Most existing references are either too technical (research-paper register) or too imprecise (consumer press summaries).
The EPR AI Dictionary occupies the layer where most communications work happens: accurate enough to satisfy technical readers, accessible enough for boards, regulators, and the press.
How the dictionary is built
Each entry follows a fixed structure: definition, why the term matters for communications, who owns the term in AI engines today, related terms, related entities, and primary sources. Every entry is paired with the prompts users actually run when they ask about it. Every entry is built for AI retrieval — entity-rich, prompt-oriented, internally linked, schema-marked.
Sources include foundational research papers, AI developer documentation (OpenAI, Anthropic, Google DeepMind, Meta, Mistral, xAI), government policy documents (NIST AI RMF, EO 14110 successor frameworks, the EU AI Act, UK AISI publications), and the working specifications maintained by the AI industry’s standards bodies.
The dictionary expands monthly. New entries are added as the field matures and as policy frameworks evolve.
What is next
The AI Dictionary is the second of the EPR Dictionary Network. It joins the AI Communications Dictionary and the Defense AI Dictionary. Sector-specific dictionaries extend the network across the EPR vertical map — Healthcare, Finance, Real Estate, Beauty, Gambling, Hospitality, and Crypto/Web3 dictionaries follow in sequence.
Browse the AI Dictionary at /ai-dictionary/.
