FDA Issues Draft Guidance on AI Model Credibility

A risk-based credibility framework for AI models used in regulatory submissions is now open for industry comment, following FDA's release of draft guidance on January 6, 2025, the agency's first formal recommendations on AI use in drug and biological product development. For regulatory affairs leads and QA directors already embedding AI tools into submission workflows, the guidance introduces a structured methodology that will directly shape how those tools are validated and documented before reaching the agency.

Central to the framework is the concept of context of use: a defined description of how an AI model addresses a specific question of interest. Sponsors are expected to assess and establish model credibility relative to that context, then determine which credibility activities are necessary to demonstrate that the model's output is fit for regulatory decision-making on safety, effectiveness, or quality. The agency notes this approach is consistent with how its reviewers have already been evaluating submissions containing AI components, a signal that informal expectations are now being codified.

The evidentiary basis for the guidance is substantial. FDA drew on feedback from an expert workshop convened by the Duke Margolis Institute for Health Policy in December 2022, more than 800 external comments on two discussion papers published in May 2023, and its own review experience across more than 500 drug and biological product submissions with AI components since 2016. That volume of prior review activity suggests the agency has developed internal benchmarks that the draft guidance is now making explicit.

Cross-center coordination underpins the document. FDA's human and animal medical product centers developed the guidance collaboratively with the Office of Inspections and Investigations, the Oncology Center of Excellence, and the Office of Combination Products, a structure designed to ensure consistent application across product types and review divisions. For sponsors operating across multiple product categories, that alignment reduces the risk of divergent credibility expectations at different centers.

The agency is explicitly encouraging early engagement on AI credibility assessments, framing pre-submission dialogue as a mechanism for sponsors to align on context-of-use definitions before formal review. Regulatory affairs teams preparing AI-supported submissions should treat that invitation as a procedural checkpoint rather than an optional courtesy, particularly where model outputs inform primary endpoints or quality determinations.

Public comment on the draft guidance closes 90 days from publication; FDA is specifically requesting input on whether the guidance aligns with industry experience and whether existing engagement options are sufficient. The comment period represents a direct channel for manufacturers and sponsors to influence the credibility standards that will govern AI use in submissions once the guidance is finalized.

Source: FDA via What's New, Vaccines, Blood & Biologics RSS Feed, January 6, 2025.