AI Compliance for FDA-Regulated QMS: Advice from KPMG | Propulsion 2026

AI is moving faster than the regulations designed to govern it, and in regulated industries, that gap carries real consequences. One FDA warning letter has already made the stakes clear.

At Propulsion 2026, Zachary "Zak" Macht, Director of Life Sciences Advisory at KPMG, delivered a wake-up call for product and quality leaders operating in regulated environments. Grounding his talk in the FDA's first warning letter to explicitly cite AI misuse, Zak mapped the practical compliance risks that come with deploying AI in GxP processes, from model drift and hallucinations to the false security of a rubber-stamp human reviewer.

His message is urgent: the regulatory landscape is shifting in real time, the AI adoption curve is not slowing down, and companies that fail to build structured oversight frameworks now will face audit findings later. Whether you manage complaints, CAPAs, or quality workflows in Propel, this session equips you to innovate without running afoul of the agencies watching closely.

Key Takeaways

The FDA Has Spoken: AI Misuse is Now a Warning Letter Offense

The agency explicitly flagged over-reliance on AI for generating critical GxP documents, including specifications, procedures, and production controls. The issue was not the use of AI itself, but the failure to maintain meaningful human oversight and awareness of legal requirements.

Companies Need to Be Prepared for a New Brand of Audit

Inspectors and auditors are preparing to ask how companies monitor ongoing AI performance, handle unexpected model behaviors, manage change control for AI systems, and establish clear accountability for AI-influenced decisions. Organizations need documented answers before those inspectors arrive.

CSV is No Longer Enough. AI Demands Continuous Assurance.

Computer software validation was built for static systems, but AI systems learn and evolve. A continuous assurance mindset, one that validates the system and governs the underlying data throughout its full life cycle, is the model that fits how these tools actually work.

A Human in the Loop is Only as Strong as the Controls Around Them

Reviewer fatigue, inconsistency across reviewers, and a growing tendency to simply trust AI outputs are real risks. A competent, independent human reviewer must add substantive value to the process, with traceable evidence that the AI's output was genuinely evaluated, not just approved.

If Everyone Owns the Data, Everyone Owns the Risk

Because AI models are only as reliable as the data that informs them, maintaining compliant, high-quality data is not a quality team problem alone. Every function that touches product data owns a piece of the compliance posture.

GxP Principles Are Your Bridge Between Innovation and Compliance

With AI-specific regulations still being codified, companies can anchor their governance frameworks in established GxP principles: risk management, good documentation practices, and audit trail traceability. The companies that build this infrastructure now will be the ones that innovate with confidence.

Don’t just follow the shift, lead it. Watch all Propulsion 2026 sessions.

Learn More

Want to see how compliant-by-design AI shows up in the product?

Explore Propel One to learn how Propel's agentic AI is built with audit traceability, governance guardrails, and security at its core, so your teams can move fast on innovation without compromising your compliance posture.

About Propulsion 2026

‍Propulsion brings together everything product and manufacturing leaders need to know in one place. Bringing you Propel’s boldest platform reveals, critical AI and market insights from top industry analysts, and real customer results. These sessions are designed to cut through the noise so you can put what matters most to work immediately.