SuperLawyer, an AI developed by Law&Company, South Korea’s leading legal consulting app, can draft a complaint for fraud that clearly shows the victim’s emotional distress within about 25 seconds. For example, it generates a complaint stating, “The plaintiff, an elderly pensioner, lent their lifelong savings but now faces a threat to their livelihood due to fraudulent acts.” SuperLawyer also handles requests such as “Show me cases where sentences were reduced in voice phishing incidents” and “Predict the prosecutor’s cross-examination questions” without any issues.
SuperLawyer, based on large language models (LLMs) like OpenAI’s ChatGPT and Anthropic’s Claude, performs a range of tasks from case law searches and summarization to drafting documents and preparing questionnaires for interrogations. At a launch press conference on July 9 in Gangnam, Seoul, Law&Company CEO Kim Bon-hwan stated, “We aim to provide a highly reliable service to legal professionals through precise search technology and a fact-checking system.”
The integration of AI technology is revolutionizing the global legal tech industry. Previously limited to case law searches, AI now sifts through extensive legal data to quickly organize relevant precedents and handle tasks like drafting responses. The technology has advanced to a level that can replace routine tasks of lawyers. According to global market research firm Future Market Insights, the global legal tech market, valued at $29.6 billion in 2024, is expected to grow to $68.04 billion by 2034 due to the adoption of generative AI.
Legal AIs learn from categorized examples of public case law, statutes, annotations, and legal theses. They analyze user questions by breaking them down into keywords, then assign weights to the most relevant pre-learned data to provide the most suitable answers. Minimizing the provision of incorrect information, known as hallucination, is crucial. Research from Stanford University indicates that law-specific LLMs from global legal tech companies still exhibit a hallucination rate of over 17%.
South Korean legal tech companies are also launching various services using generative AI. SuperLawyer, launched on July 1, uses Retrieval-Augmented Generation (RAG) technology to secure a reliable database in advance and provide the LLM with data closely related to user questions. This method ensures the LLM has a ‘cheat sheet’ for generating accurate answers.
The volume of learning material is crucial for improving the accuracy of responses. SuperLawyer has studied 1,350 types and one million pages of legal content through the domestic legal book publisher Parkyoungsa. This allows the AI to find past cases similar to specific incidents and predict sentences by calculating the median sentence. It also provides supplementary materials such as questions to ask during interrogations.
Law&Company is also developing a Korean law-specific LLM model called Solar-Legal with domestic AI startup Upstage. They plan to create a corporate AI legal model based on this and supply it to Yoon & Yang LLC within the year.
Another domestic startup, BHSN, operates Allibee, an AI solution that streamlines legal tasks such as corporate contracts and consultations. Using Allibee, tasks like drafting contracts that used to take hours can be completed in ten minutes, and the AI can identify and suggest corrections for any legal violations. BHSN plans to expand beyond South Korea to provide AI services based on the laws and precedents of countries like Vietnam and China to global companies. LBOX, which offers case law search services, has introduced LBOX AI, a service for lawyers that reduces the time needed for legal research from over 30 minutes to 2 minutes.
Despite these advancements, conflict with the Korean Bar Association (KBA) remains a significant hurdle. DR & AJU LLC has introduced AI DR & AJU, an AI service that answers legal-related questions from the general public for free within 24 hours. However, KBA is investigating this service for violating the Attorney-at-law Act, which prohibits free legal consultations.
