Open Access

ARTICLE

Building Regulatory Confidence with Human-in-the-Loop AI in Paperless GMP Validation

Manaliben Amin^*

Pharmaceutical Manufacturing, Stevens Institute of Technology, Hoboken, NJ 07030, USA

* Corresponding Author: Manaliben Amin. Email:

(This article belongs to the Special Issue: Advances in Artificial Intelligence for Engineering and Sciences)

Journal on Artificial Intelligence 2026, 8, 1-18. https://doi.org/10.32604/jai.2026.073895

Received 28 September 2025; Accepted 04 December 2025; Issue published 07 January 2026

Abstract

Artificial intelligence (AI) is steadily making its way into pharmaceutical validation, where it promises faster documentation, smarter testing strategies, and better handling of deviations. These gains are attractive, but in a regulated environment speed is never enough. Regulators want assurance that every system is reliable, that decisions are explainable, and that human accountability remains central. This paper sets out a Human-in-the-Loop (HITL) AI approach for Computer System Validation (CSV) and Computer Software Assurance (CSA). It relies on explainable AI (XAI) tools but keeps structured human review in place, so automation can be used without creating gaps in trust or accountability. Within this framework, AI provides draft recommendations such as mapping user requirements, highlighting redundant tests, or classifying deviations while subject matter experts (SMEs) review, adjust, and approve the final outcomes. Features like rationale cards, confidence bands, and traceable recommendations give SMEs the information they need to understand and, when necessary, challenge AI outputs. The framework is also supported by governance measures. These include model risk tiering to match oversight with potential impact, periodic model challenges to detect drift, and “evidence packs” that bring together AI outputs, human decisions, and audit trails into an inspection-ready format. Taken together, these safeguards show inspectors that automation is being used responsibly, not recklessly. We validate the approach in a two-project pilot, reporting mean cycle-time reductions of 32% (95% CI: 25%–38%) and higher inter-rater agreement (κ from 0.71 → 0.85), achieved under a defined governance model that includes model-risk tiering, quarterly challenge testing for high-risk models, and inspection-ready evidence packs aligned to 21 CFR Part 11 and Annex 11. These results provide preliminary empirical support so that HITL AI can improve efficiency and reviewer consistency while preserving accountability and regulatory trust. Practical case examples demonstrate that HITL AI can shorten validation cycles by 25%–40% while also improving reviewer consistency and strengthening inspector confidence. Rather than replacing SMEs, the system frees them from repetitive work so they can focus on risk and quality the areas where their judgment adds the most value. By blending automation with accountability, HITL AI provides a path for digital transformation that regulators are more likely to accept, positioning it as both a productivity tool and a model for sustainable compliance in the years ahead.

---

Keywords

---