OpenAI and the FDA are in talks about the application of artificial intelligence in drug evaluation.
High-level employees from OpenAI have met several times with the FDA in recent weeks to discuss topics related to artificial intelligence and a project called cderGPT.
The U.S. Food and Drug Administration (FDA) has been in dialogue with OpenAI to explore the use of artificial intelligence in streamlining its drug approval processes, according to sources close to the meetings. This effort is part of a broader initiative at the FDA to incorporate technology that accelerates the time it takes for new drugs to reach the market. FDA Commissioner Marty Makary expressed his concerns on a social media platform, questioning the long duration of over ten years that typically accompanies the approval of new therapies. Recently, the FDA conducted its first scientific review assisted by artificial intelligence, marking only the beginning of this transformation.
Makary made these statements following an annual meeting of the American Hospital Association, where he highlighted the potential of artificial intelligence to facilitate the approval of novel treatments, specifically for diabetes and certain types of cancer. Although the collaboration with OpenAI was not explicitly mentioned, it is known that a small group from the company has had several meetings with the FDA and some associates of Elon Musk in recent weeks. During these meetings, a project called cderGPT was discussed, which is related to the Center for Drug Evaluation, which regulates over-the-counter and prescription medications in the United States.
Jeremy Walsh, who has been designated as the FDA's first chief artificial intelligence officer, is leading these discussions, although no contract has been signed yet. Walsh has also met with Peter Bowman-Davis, a Yale student serving as the acting head of artificial intelligence at the Department of Health and Human Services, to assess the FDA's aspirations regarding artificial intelligence.
A former FDA commissioner, Robert Califf, mentioned that the agency's review teams have been using artificial intelligence for several years. Califf indicated that it would be interesting to know in detail which parts of the review process were assisted by artificial intelligence and what that means in practice. He stated that there has always been an interest in reducing review times and a general consensus that artificial intelligence could be an ally in that goal.
However, using artificial intelligence to assist in the final reviews of medications could only accelerate a small portion of a drug development process that mostly results in failures before reaching FDA review. Rafael Rosengarten, CEO of Genialis, a precision oncology company, supports the automation of certain tasks in the drug review process but advocates for policy guidelines on the types of data used to train artificial intelligence models and acceptable performance standards.
The current FDA review process takes approximately one year, although there are mechanisms in place that allow this timeframe to be expedited for promising drugs, such as "fast track" and "breakthrough therapy" designations. An industry spokesperson, Andrew Powaleny, emphasized the importance of ensuring that drugs are reviewed in a timely and efficient manner to meet patient needs, highlighting that the use of artificial intelligence in this context must be handled carefully, prioritizing patient safety.
The FDA is also conducting research on potential applications of artificial intelligence, having announced in December 2023 a grant for a researcher to develop large language models for internal use. Additionally, OpenAI has introduced ChatGPT Gov, a self-hosted version of its chatbot designed to comply with government regulations, and is working to obtain compliance accreditations that would allow it to handle sensitive government data.
