RAI UK is hosting a workshop at the AI Standards Hub Global Summit 2026.

The Summit will bring together UK and global leaders shaping the future of safe, secure, and trustworthy AI. The two-day event will explore how robust standards and credible assurance can build confidence in AI systems, enable innovation, and accelerate the adoption of trustworthy AI.

The Global Summit 2026 will be held on 16 and 17 March as a hybrid event in Glasgow and online. In-person attendance will be by invitation and sessions will be live streamed for global accessibility.

Socio-Technical Evaluation of Agentic AI

Abstract:
Traditional AI systems typically generate outputs that humans then interpret and act on, so technical evaluation—accuracy, bias, robustness—captures much of what matters. But as AI systems become more autonomous, embedded, and influential, evaluation must expand beyond technical performance. It needs to account for the people who deploy these systems, the communities they affect, the institutional and regulatory contexts they operate within, and the assumptions embedded in data, model design, and implementation choices.

Taking this socio-technical approach calls for collaboration across disciplines and meaningful engagement with end-users and other key stakeholders, so that emerging metrics, standards, and regulation reflect real-world conditions and consequences.

This workshop will bring together experts in AI and applied AI from across the UK to examine evaluation challenges across multiple sectors, including law enforcement, healthcare, and cultural heritage. Leads from major projects will present scenarios, research questions, and early findings from their work. Interactive roundtables will then involve participants in identifying shared concerns and practical priorities, helping to shape recommendations for policy-makers, researchers, and industry.

Speakers:

• Maria Liakata
• Simone Stumpf
• Radu Calinescu
• Evdoxia Taka