FDA Partners with OpenAI to Accelerate Drug Approval Process with AI
核心洞察
The FDA has initiated discussions with OpenAI (搜索) to develop "cderGPT (搜索)," an AI system aimed at streamlining drug evaluations and reducing the decade-long approval timeline for new medications.
FDA Commissioner Marty Makary recently announced the completion of their first AI-assisted scientific review, signaling the agency's commitment to modernizing regulatory processes through artificial intelligence.
The initiative involves collaboration between OpenAI (搜索), the FDA's first-ever AI officer Jeremy Walsh, and associates from Elon Musk's Department of Government Efficiency (搜索), though experts caution that proper training and validation will be essential.
The U.S. Food and Drug Administration (FDA) has begun talks with OpenAI (搜索) to implement artificial intelligence in drug evaluations, potentially revolutionizing the regulatory approval process that currently takes approximately a decade to complete.
FDA Commissioner Marty Makary highlighted this initiative in a recent statement on social media: "Why does it take over 10 years for a new drug to come to market? Why are we not modernized with AI and other things? We've just completed our first AI-assisted scientific review for a product and that's just the beginning."
The cderGPT Initiative
According to sources familiar with the discussions, a team from OpenAI (搜索) has met multiple times with FDA representatives and two associates from Elon Musk's Department of Government Efficiency (搜索) (DOGE) in recent weeks. The project, reportedly called "cderGPT (搜索)" – likely referencing the FDA's Center for Drug Evaluation and Research (搜索) (CDER) – aims to leverage artificial intelligence to expedite the evaluation of prescription and over-the-counter medications.
Jeremy Walsh, recently appointed as the FDA's first-ever AI officer, has been leading these discussions. The agency has also engaged with Peter Bowman-Davis, the acting chief AI officer at the Department of Health and Human Services (搜索), who is currently on leave from Yale University and affiliated with Andreessen Horowitz's American Dynamism team.
While no formal contract has been signed between the parties, the discussions signal a significant shift in how the agency plans to approach regulatory reviews.
Potential Impact on Drug Development Timeline
The traditional drug approval process in the United States is notoriously lengthy, with the FDA review itself typically taking about a year. However, this represents only a fraction of the overall development timeline, as most drug candidates fail before reaching the FDA review stage.
The agency already employs several mechanisms to expedite reviews for promising treatments, including fast track designation for drugs addressing serious conditions with unmet medical needs, and breakthrough therapy designation for candidates offering substantial benefits over existing options.
Robert Califf, who served as FDA commissioner from 2016 to 2017 and again from 2022 through January 2024, noted that the agency's review teams have been using AI for several years. "It will be interesting to hear the details of which parts of the review were 'AI assisted' and what that means," Califf stated. "There has always been a quest to shorten review times and a broad consensus that AI could help."
Practical Applications and Industry Response
Experts suggest that AI could immediately address certain "low-hanging fruit" in the review process. Rafael Rosengarten, CEO of precision oncology company Genialis (搜索) and a board member of the Alliance for AI in Healthcare (搜索), points to application completeness checks as an example where AI could expedite feedback to submitters.
"These machines are incredibly adept at learning information, but they have to be trained in a way so they're learning what we want them to learn," Rosengarten explained, emphasizing the need for clear policy guidance around data used to train AI models and acceptable performance standards.
The pharmaceutical industry has cautiously welcomed the initiative. Andrew Powaleny, a spokesperson for industry group PhRMA (搜索), stated: "Ensuring medicines can be reviewed for safety and effectiveness in a timely manner to address patient needs is critical. While AI is still developing, harnessing it requires a thoughtful and risk-based approach with patients at the center."
Technical Infrastructure and Challenges
In January, OpenAI (搜索) announced ChatGPT Gov, a self-hosted version of its chatbot designed to comply with government regulations. The company is also working toward obtaining FedRAMP moderate and high accreditations for ChatGPT Enterprise, which would allow it to handle sensitive government data.
However, some former FDA employees have expressed concerns about the reliability of AI models for regulatory review tasks, particularly given their propensity to generate convincing but potentially fabricated information.
The FDA has been preparing for broader AI implementation, advertising a fellowship in December 2023 for researchers to develop large language models for internal use, specifically mentioning "applications of LLMs for precision medicine, drug development and regulatory science."
Broader AI Strategy at FDA
Commissioner Makary's comments at the American Hospital Association's annual meeting earlier this week highlighted AI's potential to aid in approving new treatments for diabetes (搜索) and certain types of cancer (搜索), suggesting the agency is considering multiple applications across different therapeutic areas.
While the current focus appears to be on using AI to assist in the final review stages, former commissioner Califf suggested that "final reviews for approval are only one part of a much larger opportunity," indicating that the agency may be considering more comprehensive applications of AI throughout the regulatory process.
As these discussions progress, the FDA's approach to implementing AI in drug evaluations could serve as a model for other regulatory agencies worldwide, potentially transforming how new medications reach patients globally.