FDA announces generative AI pilot and rapid rollout plans

 In a surprise post this afternoon, new FDA Commissioner Dr Martin Makary announced a completed pilot at the FDA of a generative AI-based system for speeding up drug review - and plans to roll out the system across the agency by the end of June.

“I was blown away by the success of our first AI-assisted scientific review pilot," Makary said in a statement. "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 announcement contained very little information about the technology or the pilot. It said the rollout will be overseen by Jeremy Walsh, who just joined the agency days ago as its first chief AI officer. Walsh comes from government contractor Booz Allen Hamilton, where he was chief technologist for 14 years, including working with the FDA on contracts.

WIRED Magazine uncovered more information in a scoop published yesterday, citing unnamed sources to report that a small team from OpenAI met with the FDA and two individuals involved with the Department of Government Efficiency (DOGE) in recent weeks, with Walsh involved in the discussions. They have referred to the project as "cderGPT", likely a reference to the Center of Drug Evaluation at the FDA.

While the industry will no doubt welcome faster drug review times, the announcement raises questions about patient safety. Generative AI is notoriously error-prone and the rollout timeline is blisteringly fast - ironic for an agency that has been so hesitant to trust drug data that comes from AI tools. 

Makary justified the reduced timeline by appealing to the urgency of supporting innovation.

"There have been years of talk about AI capabilities in frameworks, conferences, and panels, but we cannot afford to keep talking," he said. "It is time to take action. The opportunity to reduce tasks that once took days to just minutes is too important to delay."

The urgency is also likely driven by the recently-announced layoffs at the FDA, which have sparked worries about delayed drug approval timelines. At AAPS National Biotech conference in Boston this week, Dr Alex Cadotte, a former FDA employee who now works for MCRA, discussed the effect of those layoffs on approval timelines.

"I'm aware of recent and planned RIFs, at least through channels that are basically available to the public, and basically the message there is that, in general, these are aimed at support personnel, rather than review staff," he explained. "So the immediate consequences of that may not be seen for a while, but it may be seen more long term. But for the most part, if you've got an application with FDA, the folks that are reviewing those are not likely to be affected."

Ironically, many of the staff who have already been laid off from the FDA were experts in AI modelling and machine learning, Cadotte said.

"With the February 14th layoffs, 75% of those [AI experts] were lost permanently," he said. "Many of them were invited back and declined. This is simply because they go out to industry and they find more lucrative positions and more stable positions. I think it's an obvious statement that that's a big loss for all of us in the sense that FDA's capacity to learn about these models is a little questionable at this point." 

https://pharmaphorum.com/news/fda-announces-generative-ai-pilot-and-rapid-rollout-plans