AI on Trial

Before AI judges us, we must judge AI. Artificial intelligence now decides who receives bail, scores the likelihood of re-offending, analyses surveillance footage, flags anomalies in financial records, predicts where crimes will occur and drafts the arguments attorneys present in court. In some jurisdictions an AI-generated risk score carries more practical weight in sentencing than anything defence counsel puts before the bench. AI on Trial is the practitioner-level guide to what that means operationally, legally and in terms of human risk.Written by Professor Kai London, a technology governance practitioner who has sat on governance committees, reviewed vendor contracts for AI systems in sensitive environments and watched organisations adopt machine learning without frameworks for accountability, the book introduces the SCALES framework for judicial AI resilience and the Evidence Chain Model: if it cannot be evidenced, it cannot be defended.Case studies are drawn from public court records, regulatory filings and verified reporting, from Mata v. Avianca and fabricated citations to risk-assessment tools in sentencing, and are traced back to the systemic failures that made them possible: missing verification protocols, the normalisation of trusting AI outputs and the regulatory vacuum in which they operated.The Premium Edition places these cases against the EU AI Act, which classifies AI in the administration of justice as high-risk, and the emerging obligations for transparency, explainability and human oversight. It is written for judges, lawyers, court administrators, legislators, regulators, technology leaders, compliance officers and anyone who builds, buys or is judged by an algorithm.Key topics include: AI in criminal and civil justice, algorithmic risk assessment, AI-generated evidence and hallucinated citations, explainability and due process, the SCALES framework, the Evidence Chain Model, AI vendor governance, the EU AI Act and high-risk systems, human oversight, liability and accountability for AI decisions.