Aftertalks #7 - Can AI Deliver Justice? Legal AI System Analysis | AI-swers

Can artificial intelligence make legal systems faster and more accessible without compromising justice? In this AI-swers Aftertalks episode, we analyze how ChatGPT, Claude, Gemini, DeepSeek, Llama, and Grok evaluated the benefits and risks of AI-powered legal systems. Legal AI could expand access to justice by supporting legal research, contract analysis, document review, precedent discovery, and case preparation. But a system optimized to produce plausible language can also invent citations, reproduce historical bias, conceal its reasoning inside a black box, and influence decisions that affect people’s freedom, property, families, and livelihoods. The discussion explores Retrieval-Augmented Generation, legal databases, explainability, audit logs, confidence scores, algorithmic bias, professional responsibility, and the limits of “human-in-the-loop” oversight. AI can assist legal professionals—but accountability cannot be delegated to a statistical model. CHAPTERS 00:00 Would you accept an AI-generated judgment? 01:30 The six-model Legal AI experiment 02:20 Do the benefits outweigh the risks? 03:00 Why AI models generally support Legal AI 04:00 Democratizing access to justice 05:20 Legal research and document analysis 06:00 Why hallucinations are dangerous in law 07:30 How the models’ approaches differ 09:15 Historical bias disguised as objectivity 10:00 The Legal AI black-box problem 12:00 Audit logs, citations and confidence scores 13:30 Who is accountable for an AI-generated error? 15:00 The illusion of human-in-the-loop oversight 16:00 Why justice requires friction 17:15 AI as an assistant—not a judge 18:00 The final question for the future of justice