Speculation about what regulators might do no longer drives the legal risk around artificial intelligence (“AI”) in life sciences. Risk is increasingly driven by the information required by regulators: a documented, risk-based assessment showing that an AI model is credible for its specific job in a drug program and how its sponsor will maintain the AI’s performance over time.
In January, the U.S. Food and Drug Administration (“FDA”) released draft guidance on using AI to support regulatory decision-making for drugs and biologics, built around a seven-step “credibility assessment framework” tied to a defined “context of use.” The framework requires sponsors to define the question of interest the AI model addresses, specify the model’s role and scope, assess model risk based on both model influence and decision consequence, and then develop, execute, and document a credibility plan commensurate with that risk. More recently, FDA’s drug center publicly emphasized that it has reviewed hundreds of submissions involving AI components across nonclinical, clinical, postmarketing, and manufacturing 2phases—with more than 100 AI-referencing submissions in a single recent year—and expects the pace to accelerate.
For drug developers, these moves will have real consequences: AI claims will increasingly be judged by whether the company can show its model was fit-for-purpose, governed, monitored, and explainable in the ways FDA now says it expects. And plaintiffs, counterparties, and whistleblowers will use the same record to press their own narratives.
FDA’s draft guidance for drugs and biologics focuses on AI used to generate information that supports regulatory decisions about safety, effectiveness, or quality, which covers a wide swath of contemporary development activities: trial design and endpoint assessment; model-informed analyses; real-world data analytics; pharmacovigilance signal detection; and manufacturing quality control. Where those AI-generated outputs affect participant safety, the reliability of study results, or product quality, they can become central to approval, labeling, and post market obligations and, therefore central to litigation risk.
To read the full article, please click here.
"AI in Drug Development Has Entered Its “Evidence” Era—and That’s Where the Legal Risk Lives," by Terry M. Henry and Naomi Zwillenberg was published in the July and August 2026 edition of For The Defense.