Legal and Governance Issues in Non-Medical Diagnostic AI Systems: From Expert Systems to LLMs

Diagnostic expert systems (DXPS) have been used for decades to emulate human problem-solving in specialized domains. This survey examines the evolution of non-medical DXPS from early rule-based expert systems to contemporary artificial intelligence approaches, including machine learning models and large language models (LLMs) for diagnostic tasks. We analyze how the shift from deterministic expert systems to data-driven AI has introduced new legal and governance challenges. Key issues reviewed include accountability and liability for automated decisions, regulatory and compliance frameworks for high-stakes diagnostics in fields like engineering and finance, and ethical considerations such as transparency, bias, and user trust. By comparing historical precedents (e.g., how legacy expert systems were validated and regulated) with emerging concerns around modern AI-driven diagnostics, the survey highlights gaps in current governance. The paper also discusses proposed policy responses and frameworks for responsible deployment of diagnostic AI outside the medical domain. Our findings underscore the need for updated governance structures to ensure accountability and safety in the next generation of diagnostic systems.