Generative artificial intelligence has found a new foothold in the federal government. On Tuesday, the U.S. Food and Drug Administration (FDA) announced the early deployment of its proprietary generative AI tool, “Elsa”—a move the agency hopes will significantly enhance internal operations and scientific workflows.
What Is Elsa?
Elsa is a large language model designed to assist FDA employees with a broad range of tasks, including scientific review, code generation, and routine administrative functions. Originally scheduled to launch by June 30, Elsa debuted ahead of schedule and under budget, according to the FDA.
Read More: Google’s Veo 3 Sparks Frustration Amid AI Glitches
The AI tool is housed within Amazon Web Services’ GovCloud, a secure platform designed for sensitive and classified data. The agency was quick to clarify that Elsa was not trained on any “data submitted by regulated industry,” thereby safeguarding proprietary research and confidential materials.
Practical Applications
Elsa’s capabilities extend beyond simple summarization. According to the FDA, the tool is already accelerating clinical protocol reviews, improving scientific evaluations, and helping to prioritize inspection targets. It can read and generate content, summarize adverse event reports, and even create nonclinical code.
One FDA scientist, Jinzhong Liu, reported that tasks which previously required several days were completed by Elsa in just minutes.
Internal Reception
FDA Chief AI Officer Jeremy Walsh heralded Elsa’s release as a transformative moment for the agency. “Today marks the dawn of the AI era at the FDA,” Walsh said. “AI is no longer a distant promise but a dynamic force enhancing and optimizing the performance and potential of every employee.”
In a separate press release in May, the FDA announced the successful completion of its first AI-assisted scientific review. Peter Marks, Director of the Center for Biologics Evaluation and Research, praised Elsa’s capabilities, saying he was “blown away” by the results. “We need to value our scientists’ time and reduce the amount of non-productive busywork that has historically consumed much of the review process,” he added.
A Cautious Optimism
Despite the promising debut, concerns remain about the broader implications of AI adoption in sensitive federal processes. One major issue is the risk of “hallucinations”—false or misleading outputs generated by AI systems. These errors are often the result of biased training data or insufficient fact-checking safeguards.
According to federal IT consultancy IT Veterans, even advanced models with built-in protections require vigilant human oversight to ensure reliability. “Human review remains essential to mitigate the risks and ensure the reliability of AI-integrated federal data streams,” the group cautioned.
Workforce Implications
The timing of Elsa’s deployment has also raised eyebrows, given the FDA’s recent workforce reductions. In early April, the agency laid off approximately 3,500 employees, including scientific and inspection staff. Although some of those layoffs were later reversed, the incident has sparked concerns about relying more heavily on automation in the face of dwindling human resources.
Looking Ahead
The FDA plans to gradually expand Elsa’s role within the agency, incorporating more advanced data-processing and generative capabilities. As the technology matures, its integration is expected to further support the agency’s mission of safeguarding public health.
Whether Elsa lives up to its promise remains to be seen. But for now, the FDA has taken a bold step into the AI era—one that could reshape the way federal agencies manage data, science, and operational efficiency.
Frequently Asked Questions
What is Elsa?
Elsa is the FDA’s in-house generative artificial intelligence tool. It’s a large language model (LLM) designed to assist agency staff with tasks ranging from scientific document reviews and summarization to code generation and operational support.
Why did the FDA develop Elsa?
The FDA created Elsa to improve internal efficiency, reduce administrative burdens on scientific staff, and accelerate processes like clinical protocol reviews and data analysis. It’s part of a broader effort to modernize the agency’s technological capabilities.
Was Elsa launched ahead of schedule?
Yes. Elsa was originally slated for release by June 30, but it launched earlier than expected and came in under budget, according to an official FDA statement.
What kind of data was used to train Elsa?
While the FDA has not disclosed full details of the training data, it emphasized that no data submitted by regulated industry was used. This is intended to protect proprietary and sensitive information.
Where is Elsa hosted?
Elsa is deployed on AWS GovCloud, a secure cloud infrastructure built to handle sensitive and classified federal data.
Are there risks associated with using Elsa?
Yes. Like all generative AI systems, Elsa is susceptible to hallucinations—inaccurate or misleading outputs. The FDA has emphasized the importance of human oversight and quality control to mitigate these risks.
Conclusion
The FDA’s early rollout of Elsa marks a significant milestone in the federal government’s adoption of generative AI. By integrating a purpose-built language model into its operations, the agency is aiming to streamline complex workflows, accelerate scientific evaluations, and reduce administrative burdens on its workforce.
While Elsa has already demonstrated promising results, its long-term success will depend on ongoing oversight, responsible governance, and transparency in deployment. As with any emerging technology—particularly one used in public health and safety—balancing innovation with caution is critical.
