Large Language Models & Generative AI in Legal
LLMs power a new generation of legal AI tools — Harvey, CoCounsel, Lexis+ AI — that draft documents, answer research questions, and analyze legal issues conversationally.
Industries Distribution
What is AI Large Language Models & Generative AI in Legal?
Large language models have catalyzed the most significant transformation in legal technology since the advent of online legal research. Within two years of GPT-4's release, virtually every major legal technology provider launched LLM-powered features, and several well-funded startups — Harvey AI foremost among them — built entirely new platforms around generative AI for legal work. The appeal is intuitive: LLMs can read and write legal text with a fluency that makes them natural tools for a profession that runs on language.
The legal-specific LLM ecosystem has matured rapidly. Harvey AI, founded by former O'Melveny & Myers attorney Winston Weinberg and backed by Google Ventures and Sequoia, has become the leading legal generative AI platform with partnerships across dozens of Am Law 100 firms. Thomson Reuters integrated LLMs across its product suite: CoCounsel for legal research, Practical Law AI for drafting assistance, and Westlaw AI for case analysis. LexisNexis launched Lexis+ AI with similar capabilities. Spellbook (from Rally) focuses on contract drafting, and EvenUp uses LLMs to generate demand letters for personal injury cases. Each platform addresses the fundamental challenge of legal LLMs: grounding outputs in verified legal authorities to prevent hallucination.
The implementation approaches vary significantly across the market. RAG (retrieval-augmented generation) is the dominant architecture, connecting LLMs to curated legal databases so that generated text is anchored in real cases, statutes, and regulations. Fine-tuning on legal corpora improves understanding of legal concepts and reasoning patterns. Constitutional AI and guardrail systems prevent outputs that could create ethical or malpractice risks. The most sophisticated implementations combine all three approaches, creating systems that generate fluent, accurate legal content while maintaining appropriate uncertainty when the law is unclear or evolving.
What Large Language Models & Generative AI Delivers
- Draft legal documents — briefs, memos, contracts, correspondence — in minutes with quality approaching experienced attorney output
- Answer complex legal research questions conversationally with cited authorities, eliminating hours of manual case law searching
- Summarize lengthy documents, depositions, and regulatory filings into actionable digests tailored to specific legal questions
- Analyze contract language for risks and inconsistencies using contextual understanding that goes beyond keyword matching
- Enable self-service legal tools that let business users handle routine legal questions without engaging the legal department
Large Language Models & Generative AI: Common Questions
Legal AI platforms use multiple strategies to minimize hallucination. RAG (retrieval-augmented generation) grounds LLM outputs in verified legal databases — when CoCounsel cites a case, it retrieves the actual case from Westlaw rather than generating text from memory. Citation verification systems check every legal reference against authoritative databases before presenting results. Confidence scoring flags outputs where the model is uncertain. And some platforms restrict generative capabilities to domains where they have verified training data, declining to answer questions outside their knowledge boundaries. These approaches don't eliminate hallucination entirely but reduce it to manageable levels with attorney oversight.
AI Roadmap — Associates spending the week on doc review. Contracts taking two weeks to turn. Both have AI fixes. We'll show you which, in what order. 72 hours.