04/07/2026
Photo By Aaron Burden on Unsplash
On June 2, 2025, the U.S. Food and Drug Administration (FDA) announced the launch of their Artificial Intelligence (AI) Tool: Elsa. AI has been seen in many parts of the clinical research ecosystem in recent years such as trial and protocol design, participant recruitment, and safety monitoring, all of which have the potential to enhance trial efficiency and safety. Earlier that year we saw the FDA release their first draft guideline on: Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products that provided a framework on how sponsors and other stakeholders could apply AI to their trials without jeopardizing participant’s safety and rights or data integrity. The FDA is also anticipating that trials using AI will eventually (if not already) start assembling trial data for regulatory and other document submissions. Using AI this way has the benefit of reducing expenses for sponsors; however, the AI’s output should always be verified by experienced staff for accuracy in case of a mistake. All the points mentioned above primarily deal with how sponsors are using AI in trials, however, it should come as no surprise that regulators will also turn to this tool in order to enhance their trial oversight and inspections.
The FDA’s Chief AI Officer Jeremy Walsh called the agency-wide rollout of the AI tool Elsa “the dawn of the AI era at the FDA”. Elsa is a large language model-powered (LLM) AI tool capable of reading, writing, and summarizing information, like OpenAI’s ChatGPT. Now it is being used for protocol reviews and scientific evaluations of submitted documents from trials but most important in the context of inspections is Elsa’s ability to identify high-priority inspection targets. By reviewing all the information submitted by a trial such as adverse events, previous inspections results, and other relevant external data, Elsa can detect subtle patterns or trends that would suggest potential risks to a trial’s participants or data and direct the FDA’s inspectional resources more efficiently. These risk evaluations can be conducted and updated in real time as it takes little time for the function to be completed. To safeguard Elsa’s functions from external manipulation, it does not learn from information submitted from regulated industry. All documents that are uploaded to the ELSA platform are under high security and protection; only authorized FDA personnel will be able to access and review these documents.
There has been lots of buzz surrounding the use of AI in the field of clinical research. Keep an eye out next week for our part 2/3 of this blog series about Agentic and Generative AI, how it affects inspection readiness, and how you can prepare for an AI assisted FDA inspection.
-The Clinical Pathways Team
Enjoy this blog? Please like, comment, and share with your contacts.