FDA Plans to Use AI to Streamline Approval Processes

The FDA plans to use generative artificial intelligence (AI) to streamline the approval process, according to an article by Commissioner Martin A. Makary, MD, MPH, and Center for Biologics Evaluation and Research (CBER) Director Vinayak Kashyap Prasad, MD, MPH, published in JAMA.

Generative AI may be able to make an initial review of documentation submitted as part of the application, which can exceed 500,000 pages, and assist in the generation of standardized tables. To support AI development, the agency plans to change its approach to the technology.¹

“The FDA will also modernize how it reviews AI-based technologies. A diversity of use cases and rapid changes in technology make the legacy review mechanism appear byzantine,” the authors write. “Rethinking our approach to AI, balancing safety and accuracy while fueling innovation, is a leading FDA priority.”

On May 8, 2025, the agency implemented an AI-assisted scientific review pilot, with plans to scale the use of the technology across all FDA centers by the end of June. The tool is intended to allow experts to spend less time on repetitive tasks which can slow down the review process. While all centers will be operating on a common generative AI system integrated into internal data platforms by June 30, the system will be tailored to fit the needs of each center after that date. Further changes will focus on expanding use cases and improving functionality.²

The tool, known as Elsa, was launched June 2, 2025. Elsa was built using the GovCloud system, allowing users to access internal documents. The tool is able to summarize adverse events, perform label comparisons, and generate code for non-clinical application databases. The agency plans to push its AI use beyond Elsa, eventually integrating the technology into processes including data processing.³

The FDA’s AI vision was outlined in an opinion piece titled, “Priorities for a New FDA.” Other subjects touched upon in the piece include healthier food for children, and speeding up the review process for life-threatening diseases, among other topics.

Beyond increasing the efficiency of the approval process, the FDA has created a roadmap to use AI-based computational models to predict toxicity-leveraging chip technology to reduce the need for animal testing. There is a possibility AI could better predict toxicity in humans, while also reducing waste and speeding-up drug development.