FDA Policy/General

FDA Elsa Rollout a ‘Nuanced Dance’: Consultant

David McFarland

— 1 min read
Share

The FDA rollout of its generative artificial intelligence (AI) staff tool Elsa is in many ways a “nuanced dance between innovation and responsibility.” That’s the view of ERB Lab Services principal consultant and owner Erika Roberts, based on her more than 15 years of experience working in various areas of the pharmaceutical/biotech manufacturing environments.

Writing for Med Device Online, Roberts says FDA is heralding Elsa as a significant step toward modernizing agency operations and enhancing efficiency, a goal she believes everyone can support.

“The official narrative from FDA press releases paints a picture of a secure, Internet-facing tool, built within a high-security GovCloud environment,” Roberts writes. “This crucial detail ensures that all information remains strictly within the agency and, importantly, that the models do not train on sensitive data submitted by regulated industries. This commitment to data privacy is paramount, especially in a sector where proprietary information is king and confidentiality is non-negotiable.”

According to the post, FDA says Elsa is already tackling a range of tasks:

 

  • accelerating clinical protocol reviews;
  • shortening scientific evaluations;
  • identifying high-priority inspection targets;
  • summarizing adverse events;
  • speeding label comparisons; and
  • generating code for non-clinical applications.

 

Roberts cautions that “while the excitement surrounding Elsa’s potential is palpable, some internal feedback suggests that there are areas for refinement and careful implementation.”

She says the agency has acknowledged that Elsa is only the first step in FDA’s overall AI journey. It is seen as a foundation upon which future agency technology advancements will be built. She says such a forward-looking perspective is crucial because the true power of AI in regulatory science “will emerge not from a single tool but from a comprehensive ecosystem of interconnected intelligent systems.”

Such an ambitious vision, she says, will depend on continuous learning and adaptations, explainability and transparency, and the irreplaceable human element.

Roberts concludes that, like a newly-trained scientist in a lab, Elsa will require “ongoing experimentation, meticulous data validation, and continuous feedback from its human collaborators. The agency’s commitment to security, as well as its plan to integrate more AI in different processes as Elsa matures, are positive indications. The real success of Elsa won’t just be measured by the speed of its initial rollout, but by its sustained accuracy, its ability to genuinely empower FDA employees, and the trust it builds within the agency and among the public.”

Read more