FDA Launches Internal AI Tool "Elsa" to Streamline Regulatory and Scientific Workflows

The U.S. Food and Drug Administration has deployed Elsa, an internal generative AI tool designed to assist employees with tasks such as protocol reviews, adverse event summarization, label comparisons, and internal database support. The tool was launched on June 2, 2025—nearly a month ahead of the agency’s June 30 deadline—and is now accessible across all FDA centers.

Built in a GovCloud environment, Elsa allegedly does not train on any regulated industry data. Its launch follows a successful pilot with scientific reviewers, where task durations reportedly dropped from days to minutes. According to Commissioner Marty Makary, the rollout was completed "ahead of schedule and under budget". Chief AI Officer Jeremy Walsh called it the beginning of "the AI era at the FDA".

Elsa is part of a broader agency modernization effort. In May, the FDA issued an internal mandate to accelerate AI deployment, and in January, it released draft guidance on AI use in regulatory decision-making. That draft outlined a risk-based framework for evaluating AI models’ credibility, lifecycle management, and transparency—but excluded operational tools like Elsa from its scope.

The launch also follows the FDA’s April announcement of its plan to replace many animal testing requirements with New Approach Methods (NAMs), which include AI-based toxicity simulations and human-derived tissue models.

While Elsa is not intended for use in regulatory submissions, it supports internal operations like clinical evaluation, inspection prioritization, and code generation for nonclinical tools. As the agency continues refining its AI tools, further integration into center-specific workflows is expected later this year.

Topics: AI & Digital