Quick Highlights
- The FDA has deployed agentic AI capabilities agency-wide to help staff plan, reason, and execute multi-step tasks.
- More than 70% of employees are already voluntarily using the agency’s LLM-based tool, Elsa.
- The new AI functions will support meeting management, pre-market reviews, review validation, post-market surveillance, inspections and compliance, and administrative workflows.
- The FDA has launched a two-month “Agentic AI Challenge,” with outcomes to be showcased at the 2026 FDA Scientific Computing Day.
- The system operates within the high-security GovCloud environment and does not train on employee inputs or data submitted by regulated industries.
2Firsts, December 2, 2025 — According to an announcement on the U.S. Food and Drug Administration (FDA) website, the agency has rolled out agentic artificial intelligence (agentic AI) capabilities across all departments, offering employees an optional AI tool to support more complex, multi-step workflows.
Agentic AI refers to advanced systems capable of achieving specific objectives by planning, reasoning, and carrying out multi-step actions. These systems include built-in guidance mechanisms — including human oversight — to ensure reliable outcomes. Use of the tool is entirely voluntary for FDA staff.
FDA Commissioner Marty Makary said the agency is steadily expanding its use of AI “to put the best possible tools in the hands of reviewers, scientists, and investigators.” He described the current moment as a pivotal point for modernizing the agency through innovative technologies, adding that the new tools could significantly boost the FDA’s ability to accelerate access to new therapies and treatments.
Prior to this deployment, the FDA introduced its large-language-model–based internal tool “Elsa” in May. According to internal data, more than 70% of employees now use Elsa voluntarily. Since its launch, program teams across the agency have continuously refined the tool to better integrate it into daily workflows.
The newly deployed agentic AI capabilities will further expand the FDA’s use of artificial intelligence in more complex tasks. The agency said the system can assist with meeting management, pre-market reviews, review validation, post-market safety surveillance, inspections and compliance, along with a range of administrative functions.
As part of the rollout, the FDA has launched a two-month “Agentic AI Challenge” for internal staff. Participants will build solutions using the new capabilities, with final demonstrations scheduled for the FDA Scientific Computing Day in January 2026.
Jeremy Walsh, the FDA’s Chief AI Officer, noted that reviewers have shown “creativity and forward-thinking” in adopting AI tools. He said agentic AI would further streamline their workflows and enhance their ability to ensure the safety and efficacy of regulated products.
According to the FDA, the AI systems operate within the high-security GovCloud — the U.S. government’s dedicated cloud platform. The models do not train on employee inputs or data submitted by regulated industries, ensuring that sensitive research and regulatory information handled by agency staff remains protected. The FDA sees this deployment as a key step in its broader AI strategy to more deeply integrate artificial intelligence into daily operations and raise overall efficiency.
We welcome news tips, article submissions, interview requests, or comments on this piece.
Please contact us at info@2firsts.com, or reach out to Alan Zhao, CEO of 2Firsts, on LinkedIn
Notice
1. This article is intended solely for professional research purposes related to industry, technology, and policy. Any references to brands or products are made purely for objective description and do not constitute any form of endorsement, recommendation, or promotion by 2Firsts.
2. The use of nicotine-containing products — including, but not limited to, cigarettes, e-cigarettes, nicotine pouchand heated tobacco products — carries significant health risks. Users are responsible for complying with all applicable laws and regulations in their respective jurisdictions.
3. This article is not intended to serve as the basis for any investment decisions or financial advice. 2Firsts assumes no direct or indirect liability for any inaccuracies or errors in the content.
4. Access to this article is strictly prohibited for individuals below the legal age in their jurisdiction.
Copyright
This article is either an original work created by 2Firsts or a reproduction from third-party sources with proper attribution. All copyrights and usage rights belong to 2Firsts or the original content provider. Unauthorized reproduction, distribution, or any other form of unauthorized use by any individual or organization is strictly prohibited. Violators will be held legally accountable.
For copyright-related inquiries, please contact: info@2firsts.com
AI Assistance Disclaimer
This article may have been enhanced using AI tools to improve translation and editorial efficiency. However, due to technical limitations, inaccuracies may occur. Readers are encouraged to refer to the cited sources for the most accurate information.
We welcome any corrections or feedback. Please contact us at: info@2firsts.com
