The US Food and Drug Administration has announced the rollout of a large language model (LLM) called Elsa, aimed at helping employees with tasks such as reviewing scientific data and identifying high-priority inspection targets. The agency claims the tool will accelerate clinical protocol reviews and shorten evaluation times.
FDA Chief AI Officer Jeremy Walsh touted the launch, saying it marks the beginning of the “AI era” at the FDA. Commissioner Marty Makary also highlighted the speed of the rollout, saying it was ahead of schedule and under budget.
However, an internal test by NBC News revealed that Elsa provided incorrect summaries of FDA-approved products or public information, raising questions about its accuracy and reliability. The issues have sparked concerns among FDA staff, who describe the tool as “rushed,” “buggy,” “overhyped,” and “inaccurate.”
Source: https://arstechnica.com/health/2025/06/fda-rushed-out-agency-wide-ai-tool-its-not-going-well