FDA's AI Credibility Framework Shifts Pharma Focus to Data Ops

The Ambiguity Just Ended

Two years ago, when a pharmaceutical sponsor asked the FDA whether AI could identify patients most likely to respond to a therapy, the agency's answer was essentially "it depends." That era of regulatory uncertainty ended in January 2025 when the FDA published draft guidance introducing a seven-step, risk-based credibility framework for AI in drug development. Twelve months later, the FDA and European Medicines Agency issued joint guiding principles—the first coordinated transatlantic regulatory position on AI in pharma.

But the arrival of clear frameworks has surfaced a more complicated question: whether these rules enable genuine AI integration or codify caution that will quickly become outdated as models advance.

Why it matters

The FDA has evaluated over 800 AI-component submissions since 2016. These new frameworks distill nearly a decade of lessons about what goes wrong when sponsors deploy models without proper documentation, change control, or evidence of human oversight. For companies currently using AI in active trials, that history raises an uncomfortable question: how many existing deployments would fail to meet the new standards today?

The Back Office, Not the Bridge

Comment letters submitted by Novartis, Lilly, Bayer, and Daiichi Sankyo reveal a shared doctrine that most AI coverage has missed. These companies are not arguing for autonomous AI systems. Daiichi explicitly stated that AI should not autonomously make dose escalation or safety decisions. Every letter drew the same boundary: AI should support qualified clinical judgment, not replace it.

What sponsors want underneath that line is aggressive automation of data operations—automated reconciliation, coding consistency, discrepancy detection, and safety surveillance. The back office work, not the clinical decision-making. Lilly's position is particularly revealing: the FDA should select for the most mature oversight of AI, not the most mature AI itself.

This reframes the competitive landscape. The industry conversation has focused for three years on patient recruitment algorithms and adaptive protocol engines. But according to these comment letters, the near-term battleground is cleaner data delivered faster to humans who still make the final calls. A brilliant model with weak governance does not clear the bar. A well-governed model doing unglamorous work does.

A Structural Gap Exposed

Lilly and Bayer independently proposed a Drug Master File-style pathway that would allow AI vendors to disclose model documentation directly to the FDA, with sponsors referencing it. This mechanism signals that sponsors recognize their own AI vendors may hold documentation that sponsors themselves cannot fully produce on demand. That structural gap between deploying AI and governing it is precisely what the FDA's credibility framework is designed to expose.

The frameworks now require sponsors to specify how each AI model is used in study protocols, identify and mitigate risks, validate continuously as data environments shift, and integrate data scientists with clinical leads throughout trials rather than at handoff points. These are not aspirational principles—they are checkboxes with audit trails attached.

Who Writes the Standard

The comment letters were not passive responses to finished policy. They were active attempts to shape what "mature AI oversight" means in practice. Sponsors are submitting detailed operational models and asking regulators to ratify them. Whether the FDA adopts the DMF-style vendor pathway will signal more about the future of AI governance in trials than any principles document published so far. If adopted, it would create direct regulatory relationships between the agency and AI vendors that currently exist only within sponsor-vendor contracts.

The sponsors filing these letters are not asking the FDA to slow down. They are asking for a governance architecture they can actually build. The FDA's next move will determine whether the 2025-2026 frameworks become an operational foundation or a compliance floor the industry immediately negotiates around.

These details were first reported by Clinical Trial Vanguard, drawing on analysis from industry observers including Shashidhar Joshi, Amit Patel, and Silvio Galea.