The New Adventure of FDA and Elsa A.I.
A long time ago, a British mathematician name Alan Turing impacted history when he proposed the question, “Can machines think?”. His research and work including the “imitation game” later called the “Turing Test” was invented to test a machine’s ability to exhibit the human behavior of “thinking”. This introduced the concept for the possibility of machines learning from experience and altering their own programming which paved the road for today’s Artificial Intelligence (A.I.) [1] .
Is there a difference between machine learning and machine intelligence (A.I.)? Machine learning focuses on a machine learning from data while growing and changing its performance on specific tasks without needing additional programming by humans. Machine intelligence is broader and includes a larger picture of a machine possessing more human-like qualities, such as thinking and making decisions to solve problems.
On June 2, 2025, the Food and Drug Administration (FDA) released they had launched ahead of schedule (insert drum roll…) their new generative Artificial Intelligence (AI) tool, “Elsa”! “Elsa” is a secured platform for FDA staff to use internally; it’s built within a high-security GovCloud environment to safeguard the sensitive data handled and allows more efficiency while modernizing workflow to better serve the people [2].
Some areas “Elsa” will be used in include [2] :
Expediting clinical protocol reviews and reducing the overall time to complete scientific reviews (days to minutes)
Summarizing adverse events to support safety profile assessments
Conducting expedited label comparisons
Generating code to facilitate the development of databases for non-clinical applications
Helping inspectors identify high yield inspection targets
FDA states “Elsa” is secure as all information stays within the agency, and that A.I. models are not being trained on data being submitted by the industry so sensitive data and research are safeguarded. “Elsa” doesn’t make regulatory decisions, this still requires human experts, and “Elsa” should be used with appropriate oversight [3] .
With this announcement, industry reception has been mixed with some questions arising around validation and verification for algorithms developed and when it will become public. Some thoughts, such as if A.I. is being used in scientific reviews of applications and unfavorable decisions were made, this could become a topic that needs more discussion.
This could impact and potentially delay companies’ products for patient access as they would need to appeal and request more information. On a positive note, an FDA user who was a reviewer stated Elsa had been accurate with minor mistakes and helpful in getting them information quickly [3] . And with this, we continue watching and learning more about FDA and Elsa on their adventures in our industries!
Check back as we keep up with industry trends and updates!
References: