AI systems used in healthcare are governed by strict ethical principles and legal requirements, from the EU AI Act to the EU Charter of Fundamental Rights. But knowing the rules isn’t the same as knowing how to apply them. Hospitals and civil society organisations (CSOs) need concrete, operational processes they can actually put into practice, day to day, to make sure these tools are used safely and fairly.
This is exactly what the AEQUITAS AI Regulatory Model was designed to do. It gives hospitals and CSOs a practical governance framework covering the full lifecycle of AI systems, from procurement and validation to deployment, monitoring, and accountability.
The model draws inspiration from approaches that have proven effective in other high-risk sectors. It echoes the logic of pharmacovigilance, the systematic tracking of adverse drug reactions, in what AEQUITAS calls an “AI-vigilance” system: a way to report and analyse incidents involving AI-supported clinical decisions, with a particular focus on equity outcomes. It also mirrors the patient safety culture familiar from aviation-inspired approaches in hospitals, built around identifying hazards, reporting incidents, and continuously improving.
At its core, the model is grounded in the EU Charter of Fundamental Rights (2009), particularly the principles of human dignity, equality and non-discrimination, and the right to healthcare, translating these rights into operational safeguards that follow an AI system throughout its use.
The five checkpoints
Rather than treating AI governance as a single box to tick, the model structures oversight around five sequential stages, each one a checkpoint designed to catch potential equity risks before they reach patients. Throughout this process, hospitals act as the main deployers, leading decisions and operational control, while CSOs play an independent oversight role, reviewing documentation, flagging equity-related risks, and conducting audits, without controlling day-to-day clinical decisions themselves.
These stages follow the natural lifecycle of any AI system in a hospital. It starts with scope and classification, determining an AI tool’s regulatory status before it’s even procured or developed. This sets the direction for everything that follows. Next comes procurement, often the moment hospitals have the greatest leverage, as it’s where suppliers can be required to commit to addressing bias risks throughout a system’s lifecycle. At this stage, CSOs are able to review procurement documentation and flag equity concerns. The third stage, validation, means independently evaluating how a system performs, not just overall, but across different patient groups, before it’s ever used with real patients.
Once a system passes validation, the deployment stage ensures the right safeguards are in place: trained oversight staff, mechanisms for clinicians to override AI recommendations, and clear communication with patients about when these tools are being used in their care. And because AI systems aren’t static, the final stage, ongoing monitoring, auditing, and accountability, keeps watching for emerging disparities long after a system goes live, with CSOs conducting independent audits and helping escalate persistent issues.
Building a shared network across Europe
Beyond the model itself, AEQUITAS is currently establishing a European network of CSOs and public hospitals. This shared governance layer means that insights, incidents, and mitigation strategies identified in one country contribute to a growing, Europe-wide knowledge base, one that benefits patients well beyond the institution where the issue was first spotted.
Together, these tools turn abstract principles into something hospitals and CSOs can actually act on, closing the gap between good intentions and equitable, everyday care.
References:
European Union. (2009). Charter of Fundamental Rights of the European Union. Official Journal of the European Union, C 303/1. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:12012P/TXT