Maat: AI Legal Assistant for Competition Law Research
Maat has been unveiled by researchers as a ReAct agent tailored for competition law professionals engaged in legal research. Unlike broader tools such as Claude and ChatGPT or specialized legal assistants like SaulLM-7B and LegalGPT, Maat focuses specifically on competition law analysis. It employs retrieval-augmented generation (RAG) to anchor cases in official documents, offers inline citations, resorts to web searches when database resources are limited, and seeks user clarification on unclear inquiries. Developed collaboratively with competition law specialists, Maat is designed to tackle challenges related to insufficient domain knowledge, lack of citations, and the fabrication of cases. Details about the system can be found in a paper on arXiv (2605.27331).
Key facts
- Maat is a ReAct agent for competition law research.
- It uses RAG to ground cases in official sources.
- Provides rich inline citations.
- Falls back to web search when database coverage is insufficient.
- Prompts users for clarification on ambiguous queries.
- Designed iteratively with competition law experts.
- Addresses inadequacies of general assistants like Claude and ChatGPT.
- Paper available on arXiv (2605.27331).
Entities
Institutions
- arXiv