EMA AND FDA GUIDING PRINCPLES ON AI: BUILDING A RISK-BASED AI STRATEGY IN DRUG SAFETY

AI can transform pharmacovigilance if it is embedded within structured risk assessment, continuous validation, governance oversight, and human accountability. A risk-based approach is key to ensuring AI improves pharmacovigilance without compromising patient safety, trust, or compliance. 

• How do we decide how risky an AI tool is? E.g. Case processing support, signal detection, narrative generation, literature screening. 

• Start AI adoption in pharmacovigilance with low-risk applications before moving to more complex, higher-risk tasks. Which tasks would you classify as low-risk “wins”? 

• Approach begins with low-risk wins before moving into more complex territory. 

• Designing oversight & human-in-the-loop controls: which PV decisions must always include human validation?  

• What KPIs should we be tracking to know whether our AI is genuinely helping PV 

• Would you create a dedicated AI governance committee? Which functions must be represented (PV, QA, IT, Data Privacy, Legal, Clinical)