Research team led by Dr. Abdel-Karim Al-Tamimi at Sheffield Hallam University, alongside partners across Europe and beyond, have been awarded €16.6 million to lead a project developing a new generation of trustworthy generative AI for clinical decision support.
AIM-SAFE is developing a European multi-agent framework for trustworthy generative AI in clinical decision support, aimed at improving patient safety in polypharmacy management and oncology treatment optimisation by using modular, locally deployed containerised small language model (SLMs) agents that collaborate on clinical tasks such as adverse drug event prediction, deprescribing support, pharmacogenomic reasoning, and phenoconversion modeling, while operating in hospital settings with strong privacy protections, continuous federated learning, and explainable AI, all in line with European data governance rules including GDPR, the AI Act, and EHDS. Beyond model development, AIM-SAFE also equips researchers and clinicians with the knowledge, tools, and governance frameworks needed to use generative AI responsibly in practice, offering open-source toolkits, FAIR datasets, and Safety Co-Pilot modules to advance trustworthy healthcare AI. By strengthening European digital sovereignty, reducing reliance on non-EU AI providers, and ensuring compliance with EU ethical, legal, and regulatory principles, AIM-SAFE aims to make clinical decision support safer, more explainable, and more human-centred, while helping reduce adverse drug events, improve patient care, and lower mortality and healthcare costs.
Sheffield Hallam University plays a pivotal role in AIM-SAFE's trustworthiness and safety agenda. Led by Dr. Abdel-Karim Al-Tamimi, alongside Dr. Liam Humphreys and Prof. Rob Copeland, SHU takes a leading role in developing an integrated framework for bias detection, safety monitoring, and hallucination identification across the platform. Another key contribution of the SHU team is the design and delivery of a production-ready continuous monitoring toolkit, enabling real-time tracking of model performance, emergent demographic biases, and safety risks in clinical deployment. This includes implementing a clinician-in-the-loop feedback pipeline and longitudinal compliance reporting for regulators and hospital governance boards. SHU additionally contributes its wellbeing research expertise to ensure patient-centred outcomes throughout the project's clinical pilots.
This work is closely aligned with SHU's newly established Centre of Excellence for AI and Robotics (CEAIR), led by Prof. Alessandro Di Nuovo. Within CEAIR, Dr. Al-Tamimi serves as Research Theme Lead for Generative AI for Social Good, a theme that places responsible, human-centred AI innovation at its core. AIM-SAFE represents a flagship instantiation of this vision, translating cutting-edge generative AI research into meaningful, trustworthy, and equitable solutions for some of healthcare's most pressing challenges.
