Reference
Glossary
The vocabulary of legal AI, defined for practitioners.
- Agent
- An AI system that can plan and take a sequence of actions toward a goal, such as gathering documents and drafting a response.
- Context window
- The amount of text a model can consider at once — relevant to how much of a contract or record it can analyze.
- Fine-tuning
- Further training a model on specialized data, such as a firm's precedents, to improve performance on specific tasks.
- Grounding
- Tying an AI answer to verified sources so its claims can be checked and cited.
- Hallucination
- A confident but false output, such as a fabricated case citation.
- Large language model (LLM)
- A model trained on vast text that can generate and analyze language.
- Prompt
- The instruction or question given to an AI model.
- Retrieval-augmented generation (RAG)
- A technique where the system first retrieves relevant documents, then generates an answer based on them.
- Technology-assisted review (TAR)
- Machine learning used to prioritize and classify documents in discovery.
- Zero data retention
- A vendor commitment not to store prompts or outputs after processing — key for confidentiality.