IPM Take
This is a quiet but important turning point. AI is becoming part of how medicines are discovered, tested, manufactured and monitored. If regulators wait until AI is fully embedded, governance will always be late. The FDA-EMA principles do not solve every problem, but they create a shared language around context of use, risk, data governance, lifecycle management and human oversight. That is exactly the kind of infrastructure personalised medicine needs before AI becomes invisible inside evidence generation.
Executive Summary
On 14 January 2026, FDA and EMA jointly published ten guiding principles of good AI practice in drug development. The principles cover human-centric design, risk-based approaches, standards, clear context of use, multidisciplinary expertise, data governance, model design, performance assessment, lifecycle management and clear information for users. EMA says the principles guide AI use in evidence generation and monitoring across all phases of medicine development, from early research and clinical trials to manufacturing and safety monitoring.
Why it matters
- Regulators: Need common expectations so AI-supported evidence can be reviewed consistently across jurisdictions.
- Industry / innovation partners: Must document context of use, data quality, model performance, lifecycle management and human oversight from the start.
- Patients / advocates: Should expect AI-supported medicines development to be transparent, accountable and designed around safety, not only speed.
Previously, AI in drug development was often discussed as a productivity tool: faster discovery, faster protocol drafting, better modelling, improved manufacturing and stronger safety monitoring. That framing is incomplete. Once AI influences evidence used for regulatory decisions, it becomes part of the trust infrastructure behind medicines.
What has changed is the transatlantic alignment. FDA and EMA are not simply encouraging AI. They are setting principles for how it should be used responsibly across the medicine lifecycle. The emphasis on context of use is especially important because the same AI model may carry very different risk depending on whether it supports exploratory research, clinical trial decisions, manufacturing control or post-market safety analysis.
There is no single patient group. The affected population includes all patients whose future medicines may be developed, evaluated or monitored using AI-supported evidence. That makes this a cross-disease issue for oncology, rare diseases, neurology, cardiovascular disease, immunology and beyond.
For IPM, the implementation message is direct. AI in personalised medicine will not be credible because it is advanced. It will be credible when it is governed, documented, validated and understandable enough for regulators, clinicians, patients and payers to trust.
