Algorithms that learn patterns from labeled legal data — relevance decisions, risk labels, outcome records — to make predictions on new documents or cases; TAR is the most established application.
Last reviewed: 2026/05/19
Enterprise AI for portfolio-level contract analysis and institutional memory.
Move from this definition to role-based legal AI shortlists and the selection criteria that matter for each type of legal team.
Am Law 200 and global firm workflows: accuracy at scale, security compliance, and matter-level auditability.
Legal department workflows: contract lifecycle, regulatory tracking, outside counsel management, and risk.
Last reviewed: 2026/05/19. Definitions are written by the LawyerAI Editorial team. Commercial relationships are disclosed and do not determine editorial scores or conclusions. See our Sponsorship & Affiliate Disclosure.
Machine learning (ML) is the branch of artificial intelligence in which systems learn patterns from labeled training data and apply those patterns to make predictions or classifications on new, unseen inputs — without being explicitly programmed with the decision rules. In legal applications, ML models learn from annotated examples: attorney relevance decisions in document review, risk labels applied to contracts, historical litigation outcomes, or coded regulatory filings. The model learns what distinguishes relevant from non-relevant documents, high-risk from standard contracts, or successful from unsuccessful arguments, and applies those learned distinctions to new inputs. ML is the broader category encompassing both traditional approaches (random forests, support vector machines) and modern deep learning.
Machine learning underlies most legal AI tools, whether or not that is apparent to the user. Technology-assisted review, contract risk scoring, clause deviation detection, and litigation outcome prediction are all built on ML models trained on legal data. Understanding the basics of how ML works helps lawyers evaluate vendor claims, understand tool limitations, and ask better procurement questions.
The most important practical implication is data dependency. ML models learn from labeled data — and perform best on data similar to what they were trained on. A contract review model trained predominantly on U.S. commercial contracts may perform poorly on European or Asian law agreements. A TAR model trained on English-language documents may fail on Spanish-language materials. Performance claims should always be evaluated with reference to the training data distribution.
ML models also degrade over time if the distribution of inputs changes while the model remains static. A billing guideline compliance model trained on 2020 guidelines will generate errors when applied to 2025 guidelines that have changed. Models require monitoring and retraining.
Relativity applies ML in its Active Learning module for document relevance prediction and in analytics features for email threading, near-duplicate detection, and concept clustering. Luminance applies ML for contract clause classification and deviation detection, with models trained on large commercial contract datasets.
Kira is a purpose-built ML contract analysis tool that allows users to train custom models on novel clause types using a relatively small number of annotated examples — reducing the data requirements for deploying ML on specialized legal tasks.