Responsible Generative AI Governance in Enterprise and Healthcare Environments

Abstract

This study delves into key policy issues required for the regulation of Generative AI in high-stakes domains such as enterprise management and clinical health care. As more and more organizations deploy large language models in their tools and systems for automation, decision support, or patient care, questions of data privacy, algorithmic bias, and transparency in decision-making processes are crucial. A curated data set of 397 cases of AI failures and successes in different corporate and medical departments is used in this study. We are employing mixed methods (qualitative risk assessment instruments and quantitative performance measures) to analyse the effect of governance arrangements on the operational effectiveness and ethics. Our findings indicate that decentralized governance of data can lead to increased risks in managing data, and that data governance with a human-in-the-loop can significantly enhance data diagnostics and compliance to company policies. The study offers a blueprint for action for policymakers on the need to adopt effective guardrails to balance innovation with safety. The identified key bottlenecks in generative systems' integration are extracted from the technical logs and policy documentation in the dataset. Finally, it is important to emphasize that a multidisciplinary governance approach should be promoted, considering the ethical transparency and institutional accountability aspects, and a permanent and dynamic monitoring of the generative technologies should be maintained in professional contexts, due to the natural dynamism of these technologies.