Artificial Intelligence (AI) Advancement in the Drug/Biologics Regulatory Decision Making World!

On December 08, 2025, FDA Qualifies First AI Drug Development Tool, Will Be Used in 'MASH' Clinical Trials | FDA.

What does this mean for the industry?

1. Is AI a tool to you in advancing drug approvals?

2. Should you consider AI use in your space now that FDA has "qualified" the first AI tool?

3. Does AI tool need to be validated before submitting to the FDA?

It's important to note that in the Drug Development Tool Qualification Letter, the FDA emphasizes the following:

"The tool is designed to assist pathologists and is intended to complement rather than replace the consensus reading process (i.e., utilizing two pathologists and a tie-breaker, if needed) currently being used in MASH trials for endpoint evaluation. The tool should be integrated within the existing consensus reading process without substituting for any pathologist or the tie-breaker currently used in MASH trials."

I see this announcement to be a great progress in the use of AI in advancing drug development (#1 question) and that companies should take advantage of AI in their work (# 2 question). As AI continues to evolve, I believe that FDA is comfortable with using AI as a "qualified tool" as a complement to human intervention in the final decision-making process. This way, the risk is controllable...

The draft Guidance on Considerations for the Use of Artificial Intelligence To Support Regulatory Decision-Making for Drug and Biological Products supports the first AI qualification letter FDA made above.

Based on the draft Guidance and the first qualification letter, I would say that as long as you "qualify" your AI tool for your "Context of Use" to be supportive/complimentary to the human decision-making process, you have a good chance of advancing AI use in your organization with FDA blessing.

To me, at this time it is still quite difficult to fully "validate" (#3 question) AI for the following reasons: 1) hallucinations (i.e. making up answer) still happen, 2) you don't have control over AI engines (chat GPT, Google Gemini, MS CoPilot AI, etc.) unless you can build your own, 3) lack of consistent responses (same question asked may still lead to slightly different answer) even with high quality data input, and 4) the word "validate" does not even exist in the AI draft Guidance...to name a few...

There are so many uses of AI in the drug development process as shown in the Common Technical Document (CTD) Triangle below from GMP, GLP, CGP including its use the Pharmacovigilance space in the post-approval process...I would take advantage of AI and remember to have high quality input to obtain high quality output along with risk control using human intervention! (Ask yourself: would you be ok with FDA approve your product using AI alone without human assessment?)