## TL;DR
AI is entering the legal profession — from automated contract review and e-discovery to legal reasoning and regulatory compliance. While LLMs cannot replace lawyers, they dramatically accelerate document-intensive legal work and enable compliance at scale.

## Core Explanation
Legal NLP tasks: (1) Contract analysis — extract clauses, obligations, parties, dates, and flag risky provisions. Legal document types span M&A agreements, NDAs, employment contracts, and regulatory filings; (2) Legal judgment prediction — given case facts, predict judicial outcomes (controversial and jurisdiction-dependent); (3) E-discovery — search and classify millions of documents for litigation; (4) Statute retrieval — find relevant laws and precedents; (5) Legal summarization — condense long rulings into key holdings.

## Detailed Analysis
Legal-specific LLMs: LegalBERT (domain-adapted BERT), SaulLM (7B & 14B parameter legal LLMs trained on dedicated legal corpus), ChatLaw (Chinese legal assistant). Key challenges: (1) Hallucination — fabricating case citations is unacceptable; retrieval-augmented generation (RAG) with verified legal databases mitigates this; (2) Confidentiality — legal data is highly sensitive, driving demand for on-premise/fine-tuned models; (3) Jurisdictional specificity — laws vary by country, state, and court circuit. EU AI Act compliance tools automate risk classification, documentation generation, and human oversight tracking. The 2025-2026 trend: AI agents performing multi-step legal workflows (draft→review→redline→approve) with human-in-the-loop.

## Further Reading
- Stanford Legal Design Lab (AI + Access to Justice)
- Harvey AI (Legal LLM startup)
- ICAIL: International Conference on AI and Law