The FDA is moving quickly to bring artificial intelligence (AI) into the heart of its regulatory review process. On May 8, 2025, the agency announced the successful completion of its first AI-assisted scientific review—and it’s already scaling the technology across all centers.
FDA Commissioner Dr. Martin Makary has directed each product center (drugs, devices, biologics, and others) to begin immediate deployment of generative AI tools. By June 30, 2025, FDA expects all centers to be using a shared, secure AI platform to support internal operations and review workflows.
FDA Director Martin Makary speaking with the FDLI president and CEO, Christine Simmon
What the Pilot Showed
The pilot took place at the Center for Drug Evaluation and Research (CDER), where FDA reviewers used a generative AI tool to support documentation and analysis. According to Jinzhong (Jin) Liu, Deputy Director at CDER, tasks that once took three days were completed in just minutes. The tool helped summarize documents, flag inconsistencies, and support routine aspects of the scientific review—freeing up reviewers to focus on more complex, higher-risk issues.
Agency-Wide Rollout Underway
The rollout is being led by two newly appointed leaders:
Jeremy Walsh, FDA’s first-ever Chief AI Officer
Sridhar Mantha, Senior Advisor on AI and former CDER informatics lead
The AI system will be integrated with FDA’s internal data platforms and tailored to each center’s needs, while maintaining security and compliance with federal information systems standards. The agency emphasized that AI tools are meant to support—not replace—human reviewers.
Why It Matters
This effort could change how quickly applications move through FDA, particularly those that are well-organized and formatted in a way that allows AI tools to do their job effectively.
It also signals a broader cultural shift at the agency: one that embraces emerging technology to meet the demands of increasingly complex submissions and a high volume of applications.
What’s Still Unclear
FDA’s announcement raises several key questions for industry:
- Will companies be notified when AI tools are used during their review, and can they request insight into how AI influenced the process or handled their data?
- Will AI impact review timelines, decision-making, the frequency of additional information requests (e.g., identifying gaps or inconsistencies), or deficiencies?
- How is proprietary or confidential data protected, and is submission content ever used to train the FDA’s AI models or shared externally?
- Will the FDA provide guidance on how to structure submissions for AI readiness, especially for visuals (e.g., device diagrams, images) and file formats?
So far, the FDA has emphasized that AI will support—but not replace—human reviewers. As the agency scales deployment across centers, further clarity is expected soon.
What Companies Can Do Now
To prepare for this shift, companies can take proactive steps to improve the AI-readiness of their FDA submissions:
- Use structured formats and clean data – AI tools rely on consistent formatting and labeling to extract and process content efficiently. Submitting in FDA-accepted structured formats like eCTD (which uses XML) or eSTAR (for devices) helps ensure AI tools can efficiently parse and analyze your submission—improving speed and consistency in review.
- Reduce redundancy and inconsistencies – Conflicting information across documents (e.g., labeling vs. protocol vs. summary) can cause confusion for reviewers—and for AI tools. Make consistency a core QC priority.
- Organize by topic, not just chronology – Structure your submission to group related materials logically. If your file reads like a clean table of contents, it’s more likely to align with how FDA’s AI tools will review it.
- Document your assumptions – Use cover letters, summaries, or module introductions to flag key points and explain your approach. AI tools can identify and organize these, making it easier for reviewers to follow your logic.
- Stay alert for future FDA guidance – Expect the agency to issue new policies, FAQs, or submission tips specific to AI-assisted review in the coming months.
