Large Language Models in Legal Systems

A major new open-access survey published on December 24, 2025 maps how large language models (LLMs) are being used across legal systems-document drafting, case analysis, legal research, compliance monitoring, and legal education-while repeatedly warning against overstating reliability in high-stakes settings.

- 

Its central contribution is not hype, but categorization: where LLMs help, where they fail, and what evaluation resources exist to measure them.WHERE LLMS ARE BEING USED

Across the literature, the dominant use cases cluster into five buckets:

• Drafting and rewriting: first-pass clauses, standard notices, issue-spotting checklists, and plain-language restatements.

• Research support: summarizing authorities, extracting holdings, and generating search pivots-often paired with retrieval systems that pull from controlled corpora.

• Case and contract analytics: classification, timeline extraction, and identifying inconsistent definitions or missing terms.

• Compliance and policy monitoring: scanning obligations against internal policies, especially where organizations face frequent regulatory change.