Despite describing the US Food and Drug Administration’s (FDA) recently disclosed use of artificial intelligence (AI) in scientific reviews of pharmaceuticals as something that’s “needed” by industry, one QualityHub expert nevertheless questions whether such AI expansion at the agency could lead to so-called “AI poisoning,” potentially misguiding well-meaning reviewers and possibly skewing the types of medical products that reach patients.
In what it calls an “historic first” for the agency, the FDA said on May 8 that the use of generative AI (GenAI) tools means reviewers will “spend less time on tedious, repetitive tasks that often slow down the review process.” The news came following completion of a drug-focused GenAI pilot for reviewers; the AI-assisted approach will be rolled out FDA-wide, including for medical devices, by the end of June. That’s when “all centers will be operating on a common, secure, generative AI system integrated with FDA’s internal data platforms,” the agency said.
In lay terms, GenAI creates new content and data by analyzing available datasets. Many regulatory and quality professionals working in pharma and MedTech already use GenAI to help assemble new product submissions, discover product safety risks, pinpoint and identify production and quality trends, make adverse event reports, and more. (Related Story: ”QHub Q&A: FDA’s Lean Into AI for Reviews Gets Thumbs-Up From Ex-FDA Digital Expert” – QualityHub, May 20, 2025.)
An AI assist for FDA reviewers is “something that is needed,” said Joe Silvia, a QualityHub subject matter expert (SME) and head of MedWare Cyber. “Any way that the agency can make the submissions lift a little lighter is certainly welcome.” He optimistically added that using AI during the review process could lead to an increase in product submissions, allow for better consistency, and cut down on time to market.
Silvia is concerned, though, that bias – either intentional or unintentional – could make its way into algorithms and, ergo, influence the work of unassuming FDA reviewers.
“My fear of AI has always been AI poisoning and AI bias,” Silvia told QualityHub in an interview. “The people who write the AI algorithm, they’re the parents of it and they can dictate how that algorithm operates and what data is ultimately seen – and that leads to AI poisoning. There are a lot of different things that AI can do to be leveraged for the greater good, but it obviously can also be used for malicious purposes.”
Silvia says, for example, that AI bias could give an advantage to large firms. “An FDA reviewer could look at AI-generated data on a manufacturer and say, ‘Company XYZ has been around for 20 years. They have a medical device family on the market with more than 30 products, their track record submissions have been all good, and they always submit good information,’” he said. “And then boom – what happens is, that large company gets its device to the marketplace faster, whereas a small startup might go through hell to get the same clearance or approval because a small company might not have that historical data that FDA algorithms could leverage.
“So, there might be some loopholes that AI doesn’t necessarily account for,” Silvia added, noting that only time will tell how the FDA’s new initiative will ultimately unfold.
In the meantime, the FDA says it will “assess performance, gather user feedback, and refine features” that will assist the agency in advancing the public health and report back with more details sometime in June.
“I’m always optimistic about technology, and at heart I’m a technologist,” Silvia said. “So, my view is, if there are any advances that the FDA can make, then that’s great. It’s all about innovation. So, at the end of the day we must embrace AI – you can’t run from it and ultimately you must embrace it – but companies and the FDA should make sure they keep one eye wide open.”
[Editor’s Note: If your MedTech or pharmaceutical company needs assistance with FDA submissions – particularly in the era of artificial intelligence – QualityHub’s trusted quality and compliance consultants, project managers, and regulatory experts can help get your products over the finish line. Discover all we can do for you at qualityhub.com.]
