Regulatory professionals in the pharmaceutical industry are facing several challenges as the U.S. Food and Drug Administration (FDA) undergoes significant changes, including layoffs, leadership transitions, and shifts in guidance. Three experts recently discussed these issues with BioSpace, focusing on the complexities introduced by artificial intelligence (AI) and supply chain demands.
In January, both the FDA and European Medicines Agency released new guidance on AI use in drug development. The document outlines 10 principles to guide AI’s role throughout a drug’s lifecycle but lacks detailed instruction, according to Gurpreet Gill-Sangha, a former supervisory pharmaceutical scientist at the FDA who now leads GS Consulting MD.
Gill-Sangha noted that while the guidance encourages ethical and human-centered design—summed up as “The development and use of AI technologies align with ethical and human-centric values”—it remains broad. She said this is only an initial step for companies looking to use AI in presenting data to regulators. Until more comprehensive guidelines are issued, she advised companies to focus on solid foundational data when submitting applications.
“The fundamentals cannot go away,” Gill-Sangha said. “AI can support the data, but if you submit an application that’s weak on fundamentals, but it’s strong on AI, that’s not a good way to take a step forward.”
She also emphasized the importance of engaging with FDA officials before submitting applications: “Many times…that dialogue doesn’t happen, even at large pharmas,” she explained, which can lead to response letters or rejections.
Rahul Gupta, president of GATC Health in Irvine, California—a company using AI for drug discovery—highlighted another challenge: adapting fast-evolving AI systems within slower regulatory frameworks. Gupta pointed out that current FDA protocols were designed for static products rather than adaptive models.
“FDA frameworks were built to evaluate fixed products and static datasets, while AI systems learn and update constantly,” Gupta stated in written comments. He questioned whether a changing AI model could still be considered validated under existing rules.
Gupta suggested solutions such as freezing models during regulatory review periods and being transparent about their limitations. He wrote: “The practical path forward isn’t to slow innovation, but to make it governable by freezing models for regulatory use, being candid about data limits, treating AI outputs as supportive rather than decisive evidence, and engaging FDA early to align expectations before precedents harden.”
Supply chain management is another concern for regulatory professionals. Ayda Delpassand of RadioMedix described difficulties securing multiple suppliers for isotopes used in radiopharmaceuticals due to limited industry options and strict regulations. She also cited obstacles in obtaining FDA authorization for manufacturing sites since inspections often require a commercially available product submission.
“It’s very challenging to get a manufacturing site authorized by the FDA,” Delpassand said. She recommended securing supply agreements early and working with organizations already approved by the agency.
Both Gill-Sangha and Delpassand stressed clear communication with regulators as essential for successful submissions. According to Gill-Sangha: “So, the applications that are extremely clear, concise—they tell you that this is their conclusion…and it’s based on XYZ…fly through for approval.”
Delpassand encouraged viewing regulators as partners: “I never really look at regulatory as being this entity that’s trying to keep me from getting a drug across the finish line…I always look at them as a partner in this.” She added that proactive engagement is key: “You essentially want to interact so you don’t have to react…”
