Trustworthy AI Development Lifecycle

Abstract

The rapid adoption of Artificial Intelligence (AI), particularly Generative AI, introduces novel risks across technical, social, and ethical domains, ranging from data breaches through data manipulation to discrimination and more. Existing governance and risk management frameworks often address these risks in a fragmented and reactive manner, lacking holistic integration into the AI development lifecycle. This research proposes to develop a unified framework for risk identification and mitigation across all attributes of Trustworthy AI, including robustness, fairness, explainability, transparency, reproducibility, privacy, security, and accountability, embedded directly into the AI development lifecycle. Relying on the design science methodology and informed by the findings of an empirical exploratory study performed in industry, this extended abstract presents a problem statement and a research plan to address it. The envisioned research includes an extensive literature review, framework development, expert validation, and a real-world case study. The proposed framework will allow Trustworthy AI implementation across all industries as well as bridge the existing gap between research, regulation, and practice by offering actionable, consolidated practices that support proactive risk management. The framework aims to empower developers and organizations to produce AI systems that are not only innovative but also compliant, safe, and socially responsible, thereby enabling sustainable and Trustworthy AI adoption.