FDA–EMA Joint AI Guiding Principles (January 2026)

The ten principles address the governance framework for AI systems in drug development, data quality and representativeness requirements, model transparency and explainability, validation and testing standards, human oversight requirements, monitoring of deployed systems, documentation and audit trails, risk-based approaches to governance, stakeholder communication about AI use, and the obligation to update practices as science and regulation evolve.

For pharmacovigilance specifically, these principles reinforce the CIOMS WG XIV framework: AI is a tool that must be governed, validated, monitored, and overseen by qualified humans. The principles explicitly state that AI used in safety surveillance must be validated for its intended use case, maintain audit trails, and be subject to human review before output influences regulatory decisions.

The fact that FDA and EMA — historically the two agencies most likely to diverge on procedural detail — chose to publish this jointly is itself a signal worth understanding. It means a company building AI governance documentation no longer has to guess whether satisfying one regulator risks conflicting with the other on this specific topic; the ten principles function as a shared baseline both agencies will reference, even though neither is legally bound to follow them to the letter.