Webinar

Webinar: FDA Perspectives on AI in Biometrics — Key Pitfalls, Bias Risks, and What Sponsors Must Get Right

Ying Li |
Webinar: FDA Perspectives on AI in Biometrics — Key Pitfalls, Bias Risks, and What Sponsors Must Get Right

This webinar is part of AI on the Hill, a series examining how artificial intelligence is reshaping regulated healthcare and clinical development.

Central Question

As AI becomes increasingly embedded in clinical development, the industry faces a critical question: “What actually holds up under FDA review — and what doesn’t?”

Overview

The webinar brought together perspectives from the U.S. Food and Drug Administration, global biostatistics leadership, and legal experts to examine how AI is evaluated in biometrics today. Rather than exploring hypothetical scenarios, the discussion focused on practical FDA-facing realities, sponsor mistakes, and accountability requirements accompanying AI adoption in regulated environments.

Core Takeaway

AI is no longer judged as a standalone tool — it functions as part of the evidence itself. Once AI meaningfully influences data selection, statistical analysis, subgroup interpretation, or trial conclusions, it becomes subject to FDA scrutiny. This shift requires teams to prioritize accountability, regulatory defensibility, and reviewer-facing impact over mere efficiency gains.

Three Key Areas for Sponsor Success

1. Acceptable vs. Unacceptable AI Applications

FDA reviewers distinguish between exploratory AI and submission-grade evidence supporting efficacy or safety claims. AI cannot function as a clinical endpoint or replace established statistical principles.

2. Bias and Validation Challenges

Common pitfalls include inadequate subgroup validation, overreliance on historical or synthetic data, and limited transparency regarding model assumptions and failure modes.

3. Governance and Oversight

Regulatory attention increasingly focuses on system governance rather than algorithms alone. Human-in-the-loop controls, comprehensive documentation, and lifecycle auditability remain essential.

Speakers

  • Hanrui Zhang — DHT Review Lead, U.S. Food and Drug Administration
  • Yali Li — Vice President, Biostatistics & Data Management, bioMerieux
  • Ran He — Founder, THC Lawyers
  • Louise Liu — CEO, Hill Research (Moderator)

Access

The full webinar is available on Apple Podcasts.