FDA Rolls Out “Elsa" To Speed Up Drug Reviews

The US Food and Drug Administration has launched “Elsa,” an in-house generative AI platform that can accelerate clinical protocol reviews, internal regulatory workflows, and safety assessments.

According to FDA Commissioner Marty Makary, as reported by Reuters, Elsa’s rollout, which the FDA completed in June 2, was nearly a month ahead of the agency’s June 30 target and under budget.

The FDA uses a large language model (LLM) to power “Elsa” and help staff analyze applications more efficiently. It can summarize adverse events and compare drug labels. “Elsa” can also generate database codes and flag high-priority inspection targets.

Image by freepik

MyMCMedia quoted Commissioner Makary as saying these tasks previously took scientific reviewers at least 2 days (with some extending up to 5 days) to complete. With “Elsa,” such activities only take minutes.

The FDA’s launch of Elsa highlights the agency’s increasing commitment to harnessing AI in healthcare.

Addressing data security, Amazon’s AWS GovCloud hosts “Elsa.” Moreover, this AI tool doesn’t train in proprietary industry submissions, ensuring confidential information remains internal.

TechTarget noted that Jeremy Walsh, the FDA’s Chief AI Officer, views Elsa as a pivotal step into a new era of AI integration at the agency. The executive also noted that the FDA plans future improvements as staff usage feedback grows.

However, agency insiders, legal experts, and other critics warned that the accelerated deployment may have overlooked performance validation and oversight safeguards.

Still, “Elsa” remains an integral part of a broader FDA modernization strategy. Should “Elsa” deliver on its promise, drug approval times can be reduced to about a month instead of the usual 10 months.