Space Regulatory Glossary AI & Data Science

retrieval-augmented generation

/rɪˌtriːvəl ɔːɡˌmɛntɪd ˌdʒɛnəˈreɪʃən/ noun (abbr.)

1. An AI architecture that grounds a language model's output in retrieved source documents rather than relying solely on the model's training data. The system retrieves relevant passages from a corpus, then feeds them to the model alongside the user's query to produce answers traceable to specific sources.
2. In regulatory intelligence, RAG enables systems to answer questions about specific filings, rules, or docket entries by pulling the actual documents before generating a response, reducing hallucination risk compared to prompting a model from memory alone.

Read: The Trust Problem — When AI Hallucinates Regulations

Related terms

large language model A neural network trained on large text corpora that generates, summarizes, and reasons about natural language. Models… embedding A fixed-length numerical vector that represents the semantic meaning of a piece of text, image, or other data. Texts… vector search A search technique that finds items by comparing the mathematical similarity of their embedding vectors rather than… hallucination An AI model output that is factually incorrect, fabricated, or unsupported by the source data, presented with the same…

Mentioned in The Downlink

• AI Verification Engineering: Architecture for Trustworthy Regulatory AI

  RAG hallucinates at 17 to 33 percent in production legal tools. The architecture of trustworthy regulatory AI: source grounding, extraction validation, confidence scoring, and standards-aligned verification.

Browse all terms in the Space Regulatory Glossary →