A team of leading AI researchers—Saurabh Pahune (Tata Consultancy Services, USA), Dr. Zahid Akhtar (State University of New York Polytechnic Institute), Venkatesh Mandapati (FedEx), and Dr. Kamran Siddique (University of Alaska)—has released a groundbreaking study highlighting the essential role of robust AI data governance in building Large Language Models (LLMs) that people can trust.
Their research demonstrates that ethical, secure, and responsible data management not only enhances model performance but also reduces bias, prevents hallucinations, and ensures regulatory compliance. By systematically addressing these challenges, the study provides a clear roadmap for organizations aiming to deploy AI responsibly.
The study delivers actionable insights for high-stakes sectors, including healthcare, banking, supply chain management, pharmaceuticals, and cybersecurity. It explores best practices, implementation challenges, and innovative strategies for embedding strong data governance frameworks within AI systems. Through real-world case studies, the research shows how LLMs can be integrated into operational workflows while maintaining transparency, accountability, and trust.
Key contributions of the research include:
A practical framework defining six pillars of trustworthy AI: data quality, ethical norms, privacy, regulatory compliance, monitoring, and traceability.
Comprehensive evaluations of how effective data governance mitigates risks such as hallucinations, bias, data breaches, and deployment failures.
Sector-specific guidance for organizations seeking to implement LLMs responsibly while meeting regulatory and ethical standards.
Innovative methodologies promoting human-AI collaboration, establishing new benchmarks for reliable, ethical, and compliant AI deployment.
This study represents a major step forward in the practical adoption of AI governance, offering businesses a structured approach to deploying LLMs safely, ethically, and effectively.
Read the full study, featured in a Q1 journal, here: https://www.mdpi.com/2504-2289/9/6/147
