Key points:

• AI will dramatically speed up PMTA reviews by summarizing and analyzing large volumes of scientific data, reducing tasks from days to minutes.
• FDA aims to standardize and streamline evaluations across all centers, including nicotine products, potentially increasing consistency and transparency in regulatory decisions.
• CTP must fully integrate AI tools by June 30, 2025, with applications expected to include risk modeling, data validation, and prioritization of pending PMTAs.

The U.S. Food and Drug Administration has announced an ambitious plan to implement generative artificial intelligence (AI) across all its centers by mid-Summer. The decision follows a successful pilot program aimed at enhancing the efficiency of scientific reviews.

“I was blown away by the success of our first AI-assisted scientific review pilot,” said FDA Commissioner Dr. Martin A. Makary. “We need to value our scientists’ time and reduce the amount of non-productive busywork that has historically consumed much of the review process. The agency-wide deployment of these capabilities holds tremendous promise in accelerating the review time for new therapies.”

The pilot program demonstrated that generative AI tools could significantly reduce the time required for routine scientific review tasks. Jinzhong Liu, Deputy Director of the Office of Drug Evaluation Sciences within the Center for Drug Evaluation and Research (CDER), noted that tasks previously taking multiple days could now be completed in minutes using AI support.

In response to the pilot’s success, Makary has directed all FDA centers to begin immediate deployment of AI technologies, aiming for full integration by the end of June. The goal is for each center to operate on a unified and secure generative AI platform that integrates with the FDA’s internal systems. Post-integration, the FDA plans to continue refining the system by improving functionality, expanding the range of supported tasks, and adapting the technology to meet the unique demands of each center.

The FDA’s new generative AI initiative is expected to significantly streamline and accelerate the premarket tobacco product application (PMTA) review process, though the agency has not yet detailed tobacco-specific workflows. Based on the pilot and agency statements, there are several ways AI could be applied within the Center for Tobacco Products (CTP).

PMTAs often involve hundreds of thousands of pages of technical, scientific, and marketing data. Makary noted that generative AI has already reduced tasks that used to take days to minutes. In CTP, this could mean:

• Rapid summarization of complex toxicological and clinical data
• Automated cross-checking of product ingredients and emissions against known health risks
• Flagging inconsistencies or missing information across large submissions

Generative AI models can assist reviewers by standardizing how information is interpreted and reported. This could reduce human bias and variation, leading to more predictable regulatory outcomes—a major concern among smaller vape companies that have argued the FDA favors Big Tobacco in PMTA outcomes.

If implemented carefully, AI could support clearer justifications for marketing decisions, which has been a longstanding criticism of CTP—particularly around the opaque criteria used to deny over 99% of flavored vape product applications.

Evaluating whether a product is “appropriate for the protection of public health” (APPH) often involves comparisons with existing tobacco products. AI could potential help researchers:

• Model behavioral usage scenarios
• Assess population-level impact of flavored vs. unflavored products
• Analyze youth usage trends using real-time surveillance data

The agency-wide rollout will be overseen by Jeremy Walsh, the FDA’s newly appointed Chief AI Officer, and Sridhar Mantha, former Director of the Office of Business Informatics in CDER. The FDA says all centers—including CTP—must be fully onboard with the secure, unified AI platform by June 30, 2025. It’s likely that early applications will focus on supportive tasks, but long term, generative AI may become central to how tobacco products are assessed.