Keynote Lectures

Real Data, Better AI: Rebuilding Trust Before Scaling Intelligence
Paulo Nunes de Abreu, Health Data Forum, United Kingdom

From 5G to 6G: Federated AI Security at the Edge of Healthcare’s Digital Infrastructure
Victor Chang, Aston Business School, Aston University, United Kingdom

When Intelligence Becomes Delegated: How AI Is Transforming Organisations
Feng Li, University of London, United Kingdom

Abstract
Artificial Intelligence is rapidly transforming healthcare, offering unprecedented opportunities to improve diagnosis, treatment pathways, operational efficiency, and patient outcomes. However, the accelerating deployment of AI across healthcare systems is also exposing a critical structural challenge: the quality, integrity, interoperability, and governance of the data upon which AI depends. This keynote explores the evolution from the “Data First, AI Later” movement towards the broader “Real Data, Better AI” vision, highlighting why trustworthy AI cannot exist without trustworthy data foundations. Drawing from international dialogues across the Health Data Forum ecosystem, including initiatives connected to the European Health Data Space (EHDS), real-world evidence, genomics, and digital health transformation, the keynote will examine:
• Why data integrity and interoperability must precede scalable AI deployment
• The risks of technology-first healthcare innovation
• The relationship between trust, governance, transparency, and AI adoption
• Real-world implementation challenges across healthcare systems
• The importance of ethical and operational readiness for AI in healthcare
The keynote will argue that the future of healthcare AI depends not only on algorithmic sophistication, but on our collective ability to build trusted, transparent, and equitable data ecosystems capable of sustaining responsible innovation at scale.

Abstract
Healthcare networks are under sustained attack, yet most hospital cybersecurity infrastructure was never designed for the distributed, software-defined environments that 6G is now bringing into view. Open RAN disaggregates the radio access network into multi-vendor, programmable components — genuine progress, but it multiplies attack surfaces in ways that centralised intrusion detection cannot adequately cover. Routing raw patient data to a central analysis node also creates a direct conflict with GDPR and NHS Digital standards. The architecture, in short, is working against itself. Federated learning addresses this at a structural level. FedAvgVChang — an adaptive aggregation framework developed for heterogeneous, latency-sensitive clinical networks — trains models locally at each node and exchanges only encrypted parameter updates across a tiered edge–fog–cloud hierarchy. No patient data moves beyond its institutional boundary. Early results show 97% attack detection accuracy, a 55% reduction in bandwidth compared with centralised approaches, a 42% improvement in training convergence, and sub-second inference times for real-time threat response. The talk sets this technical work in a broader context: how healthcare AI systems — not just the networks carrying them — need to be protected as 6G becomes operational. Drawing on the SecureAI4Public research programme, the keynote examines what the sector needs to address before the transition arrives, and what is already within reach.

Abstract
Artificial intelligence is often discussed in terms of rapidly advancing capabilities. However, the history of general-purpose technologies suggests that their real impact depends less on what the technology can do than on how organisations adapt to deploy these capabilities effectively. This keynote examines how AI represents a shift towards delegated intelligence, where responsibility for tasks, processes, and increasingly complex decisions is shared between humans and intelligent systems. Rather than a single step change, this transition unfolds along a spectrum ranging from the augmentation of human work to the delegation of tasks, coordinated workflows, and more autonomous operational and even strategic decisions. Each stage brings distinct organisational implications, raising new questions about governance, accountability, risk management, and the design of organisational structures, processes, and decision architectures. While technical capability is advancing rapidly, adoption remains uneven because deploying AI at scale requires new forms of institutional capacity. Importantly, many of the complementary changes required extend beyond individual organisations. As with earlier technological shifts, the broader benefits depend on adjustments across industries, labour markets, regulatory frameworks, and societal expectations, where the actions of different actors are interdependent and mutually reinforcing. Drawing on research into organisational transformation and the implementation of AI across sectors, the talk examines why capability often runs ahead of deployment, the sources of friction that slow adoption, and where more sustainable advantages may emerge. It suggests that the key constraint is not the technology itself but the organisational and institutional arrangements needed to absorb it safely and productively. The talk concludes with implications for leaders and policymakers, and highlights priorities for future research on how organisations and societies adapt as intelligence becomes progressively delegated.