SECAI Core · phase 5 of 15
Source grounding
The technique of connecting an AI model's outputs to specific, retrieved, and verifiable external information sources, and constraining the model to use those sources when generating its responses
The Explain card
- Plain English
- Source grounding connects a model's outputs to specific, retrieved and verifiable external sources, and constrains the model to rely on those sources rather than its own memory when answering.
- Example
- A legal research assistant is told to answer only from the retrieved statutes and to cite the section for every claim. If the sources do not cover the question, it must say so rather than improvise a plausible-sounding clause.
- Why it matters
- Grounding makes outputs checkable. An analyst can follow the citation and confirm the claim, which turns a confident guess into evidence-backed text. Ungrounded answers are hard to audit and easy to fabricate.
- Hook
- Grounding means the model shows its receipts, and it only spends what it can prove.
Where it sits in the deck
Phase 5: Interacting with Models: Prompting, APIs, and Retrieval
With a working model and clean data, learn the interaction layer — how users and systems communicate instructions and retrieve grounded answers.
