How Regulators Are Bringing AI Into Review (Without Losing Trust) with Maria Vassileva
Episode
51 min
Read time
2 min
Topics
Productivity, Fundraising & VC, Artificial Intelligence
AI-Generated Summary
Key Takeaways
- ✓ELSA Validation Limits: FDA's internal AI assistant ELSA, deployed in early summer 2025, cannot be used in formal regulatory assessments due to documented hallucinations and false outputs. The lesson: validate the entire workflow process surrounding a model, not just the model itself, to prevent misplaced trust in partially validated tools.
- ✓Risk-Tiered AI Governance: FDA's January 2025 draft guidance establishes a risk-based credibility assessment framework where required scrutiny scales with context of use. Pair this with the Total Product Life Cycle approach and a Predetermined Change Control Plan, allowing manufacturers to pre-specify algorithm modifications and data updates without triggering full re-review.
- ✓MHRA Efficiency Benchmark: The UK's MHRA cut clinical trial approval times from 91 days to 41 days by integrating AI into regulatory review workflows. Their Innovation Office provides free, confidential regulatory advice to developers. This 55% reduction over roughly six months serves as a concrete performance target for other agencies pursuing similar reforms.
- ✓Bias Mitigation in Training Data: AI models trained on historical datasets can amplify existing racial, gender, and socioeconomic disparities in diagnosis and treatment decisions. Regulators now require developers to document data provenance, demonstrate model generalizability across patient populations, and build contextually scoped models rather than assuming bias can be fully eliminated from any single system.
- ✓Global Coordination Accelerating: PMDA Japan released a three-year AI action plan in October 2025, starting with document retrieval and translation before advancing to higher-risk applications in 2026–2027. DIA's AI Consortium currently includes six to seven regulatory agencies alongside industry and academia, building shared validation templates and taxonomies to reduce duplicated effort across jurisdictions.
What It Covers
Maria Vassileva, Chief Science and Regulatory Officer at DIA, examines how FDA and global regulators are integrating AI into drug and device review processes. Coverage spans FDA's ELSA tool deployment, validation frameworks, international agency progress, equity concerns, and the multi-stakeholder governance structures needed to sustain public trust.
Key Questions Answered
- •ELSA Validation Limits: FDA's internal AI assistant ELSA, deployed in early summer 2025, cannot be used in formal regulatory assessments due to documented hallucinations and false outputs. The lesson: validate the entire workflow process surrounding a model, not just the model itself, to prevent misplaced trust in partially validated tools.
- •Risk-Tiered AI Governance: FDA's January 2025 draft guidance establishes a risk-based credibility assessment framework where required scrutiny scales with context of use. Pair this with the Total Product Life Cycle approach and a Predetermined Change Control Plan, allowing manufacturers to pre-specify algorithm modifications and data updates without triggering full re-review.
- •MHRA Efficiency Benchmark: The UK's MHRA cut clinical trial approval times from 91 days to 41 days by integrating AI into regulatory review workflows. Their Innovation Office provides free, confidential regulatory advice to developers. This 55% reduction over roughly six months serves as a concrete performance target for other agencies pursuing similar reforms.
- •Bias Mitigation in Training Data: AI models trained on historical datasets can amplify existing racial, gender, and socioeconomic disparities in diagnosis and treatment decisions. Regulators now require developers to document data provenance, demonstrate model generalizability across patient populations, and build contextually scoped models rather than assuming bias can be fully eliminated from any single system.
- •Global Coordination Accelerating: PMDA Japan released a three-year AI action plan in October 2025, starting with document retrieval and translation before advancing to higher-risk applications in 2026–2027. DIA's AI Consortium currently includes six to seven regulatory agencies alongside industry and academia, building shared validation templates and taxonomies to reduce duplicated effort across jurisdictions.
Notable Moment
Vassileva notes that Brazil's Anvisa deployed AI specifically to analyze and qualify drug impurities — a targeted, problem-driven application distinct from the document-processing focus seen elsewhere. This narrow use case produced measurable safety improvements and illustrates how agencies can start with high-impact, bounded problems rather than broad platform deployments.
Episode Transcript
Hello, everyone. Welcome to the life science rundown, where we discuss current regulatory complexities facing the life science industry and explore innovative ways to overcome those challenges. I'm your host, Nicholas Catman, president and CEO of the FDA Group. Before we get started, here's a quick word about who we are. The FDA Group is a consulting firm that helps life science companies in the areas of regulatory submissions, audit projects, mock inspection, staff augmentation, and remediation. So if you ever find yourself in need, just head over to the feagroup.com to check us out and get in touch. So today, I am speaking with Maria Vasileva. Hello, Maria. How are you? Hi. I'm great. How are you? I'm doing excellent. Thank you so much for for joining the podcast. Before we jump in for the topic today for today, would you kindly introduce yourself? Sure. I'm the chief science and regulatory officer at DIA, and I'm based in Washington, DC. DIA is a leading global multidisciplinary life science membership association, and we drive collaboration that can speed up innovation in the space of drug, drugs, biologics devices, and diagnostics. Of course as all other non profits in this space we're really dedicated to improving patient outcomes and dream of a healthier world so I'm very excited to be here today and to talk to you since we have a lot of stakeholders including industry and regulators that we very commonly interface with. So I'm looking forward to our discussion. Okay. Excellent. So I understand. So the topic today is how are regulatory agencies, particularly the FDA, navigating the opportunities and challenges of incorporating artificial intelligence into the regulatory review process while maintaining public trust, scientific rigor, and human oversight? So, with that, that statement, the first question that I would have for you is what are the most common current applications of AI in regulatory review, and how are they streamlining workflows to agencies like the FDA? That's a great question. Of of course, just like other regulatory agencies, the FDA is increasingly integrating artificial intelligence into their regulatory review processes for all medical products. And, that includes, utilizing AI to assist scientific and safety evaluations of drug biologics and medical devices and diagnostics, but also to streamline the actual, regulatory review process itself. So because, this AI application is very new, primarily AI today is being used to automate repetitive tasks, accelerate review timelines, and reduce the administrative burden, on the FDA staff so that way they can be freed up as experts to focus on the signifies, kind of a shift to because that is a whole signifies, kind of a shift toward more efficient and data driven decision making. And actually, as you mentioned in the intro, all global regulators are now thinking about and talking and and discussing how to evaluate the level of risk, of applications of AI and take more of a tailored approach. So there's always a human in the loop, for …
Get the full transcript (6,642 words) + summary by email — free
One-time email with the complete transcript and AI summary of this episode. No account needed.
One email, no spam. We’ll also show you what SignalCast does.
You just read a 3-minute summary of a 48-minute episode.
Get The Life Science Rundown summarized like this every Monday — plus up to 2 more podcasts, free.
Pick Your Podcasts — FreeKeep Reading
More from The Life Science Rundown
QA Is Not a Gate — It’s a Technical Partner, with Dr. Andrea Bell
Sep 1 · 28 min
Eye on AI
Inside the Enterprise Browser Rebuilding Security for the AI Era | Bradon Rogers, Island
Jul 13
More from The Life Science Rundown
Managing AI Risk Across Global Quality Systems
Aug 17 · 32 min
Eye on AI
India Is Becoming an Architect of the Global AI Order | Ivana Bartoletti of Wipro
Jun 16
Books, tools, and gear mentioned in this episode
SignalCast may earn commission on purchases via these links.
Tools
- ELSABy guest
by FDA
“FDA's internal AI assistant ELSA, deployed in early summer 2025, cannot be used in formal regulatory assessments due to documented hallucinations and false outputs.”
by FDA
“Pair this with the Total Product Life Cycle approach and a Predetermined Change Control Plan, allowing manufacturers to pre-specify algorithm modifications and data updates without triggering full re-review.”
other
by FDA
“FDA's January 2025 draft guidance establishes a risk-based credibility assessment framework where required scrutiny scales with context of use. Pair this with the Total Product Life Cycle approach and a Predetermined Change Control Plan.”
- DIA AI ConsortiumBy guest
by DIA
“DIA's AI Consortium currently includes six to seven regulatory agencies alongside industry and academia, building shared validation templates and taxonomies to reduce duplicated effort across jurisdictions.”
More from The Life Science Rundown
We summarize every new episode. Want them in your inbox?
QA Is Not a Gate — It’s a Technical Partner, with Dr. Andrea Bell
Managing AI Risk Across Global Quality Systems
The Case for Engaging Regulatory Early
Why Quality and R&D Default to Conflict (and How to Fix It)
Leading Through the AI Shift with ASQ's CEO Sid Bhatnagar
Similar Episodes
Related episodes from other podcasts
Eye on AI
Jul 13
Inside the Enterprise Browser Rebuilding Security for the AI Era | Bradon Rogers, Island
Eye on AI
Jun 16
India Is Becoming an Architect of the Global AI Order | Ivana Bartoletti of Wipro
Eye on AI
Jun 4
More Customers Chose the AI Agent Than Anyone Expected | Tom Chen, Aircall
The WHOOP Podcast
Jan 28
How AI & Wearables Are Shaping The Future of Healthcare with Dr. Ami Bhatt
Equity
Dec 17
Eclipse's Jiten Behl thinks the next unicorns won't be built in software
Explore Related Topics
This podcast is featured in Best Science Podcasts (2026) — ranked and reviewed with AI summaries.
Read this week's AI & Machine Learning Podcast Insights — cross-podcast analysis updated weekly.
You're clearly into The Life Science Rundown.
Every Monday, we deliver AI summaries of the latest episodes from The Life Science Rundown and 192+ other podcasts. Free for one show.
Start My Monday DigestNo credit card · Unsubscribe anytime