A Review of Emerging Legal AI Capabilities and Their Implications for Access to Justice

Dr. Tali Režun & Anton Dobrina (2026)

Abstract
Legal services represent one of the most persistently inaccessible domains in human society, despite the law's relevance to every aspect of daily life. According to the World Justice Project, 5.1 billion people -- approximately two-thirds of the global population -- currently have at least one unmet justice need. In the United States alone, 92% of civil legal problems faced by low-income individuals receive no or inadequate help (Legal Services Corporation, 2022). This article examines the structural convergence between legal reasoning and computational logic, arguing that this convergence makes legal work uniquely well-suited to AI-assisted analysis. Drawing on benchmark data from the Vals AI LegalBench leaderboard, enterprise evaluations by Box AI, and industry reporting, we review the specific technical advances in large language models -- particularly Google's Gemini 3.1 Pro Preview (released February 2026) -- that have materially advanced legal AI performance. We present six applied use cases spanning contract analysis, regulatory navigation, due diligence, family law, commercial disputes, and real estate. We address the critical limitation of AI hallucination and explain why direct document grounding in large context windows mitigates this risk. We conclude with a practical framework for non-specialist use and an assessment of the industry and policy trajectory. The central argument is that the structural isomorphism between law and code -- both operating through formal rule systems, conditional logic, and precedent-based interpretation -- positions AI not merely as a legal tool but as a potential infrastructure for democratising access to legal understanding at civilisational scale.

Keywords: legal AI, access to justice, Gemini 3.1 Pro, LegalBench, law as code, rules as code, justice gap, AI democratisation, contract analysis, hallucination risk