February 2026 Fireside Chat
Event Summary
The Data for Policy community kicked off 2026 with its first Fireside Chat of the year, focusing on the responsible deployment of frontier AI in financial services. The session highlighted an ongoing research project led by Prof Carsten Maple and Dr Anita Khadka at the University of Warwick, which investigates how AI can be safely integrated into critical infrastructure. The project addresses key challenges including AI hallucinations, prompt sensitivity, cross-jurisdictional regulatory alignment, and practical adoption by financial institutions. Its deliverables include deployable auditing frameworks, regulatory evidence packs, and sector-wide guidance to support safe and responsible AI deployment.
Dr Anita Khadka opened the session by emphasising the rapid adoption of AI in finance, from customer support and document summarization to advice-like interactions. She outlined the risks posed by confident but incorrect outputs, missing warnings, and the potential for harm to consumers, firms, and broader financial systems. She also described the project’s three-part framework: rigorous technical evaluation of AI outputs, alignment with regulatory expectations in the UK, EU, and US, and practical adoption through co-designed guidance and taxonomies developed in collaboration with stakeholders.
The conversation highlighted the importance of human oversight. Dr Khadka stressed that AI systems must be continuously monitored, tested using realistic scenarios, and paired with interdisciplinary human expertise to mitigate over-reliance and automation bias. Prof Maple reflected on parallels in other high-stakes sectors, noting the need to balance rapid AI deployment with careful governance and robust evaluation.
Discussion highlights included a nuanced understanding of harms in financial AI, ranging from factual errors and omitted warnings to downstream consequences; the challenges of evaluating dynamic, interactive AI systems; the central role of human-AI collaboration and effective handoff procedures; and the necessity of regulatory-aligned evidence that is clear, repeatable, and auditable.
Prof Jon Crowcroft closed the session by emphasising the complexity of ensuring safe AI deployment in financial services, highlighting the value of ongoing dialogue among academia, industry, and regulators. He noted that insights from this work could be applied to other critical sectors such as health and defense.
This Fireside Chat demonstrated that responsible AI in financial services requires a combination of rigorous technical evaluation, clear regulatory alignment, and collaborative adoption frameworks, ensuring AI is deployed safely, effectively, and in a way that maintains trust and compliance.
We look forward to the next session in our Fireside Chat series, where we will continue to explore critical issues shaping the future of data governance, AI regulation, and policy innovation.
About the Speakers:
Carsten Maple is the Director of the NCSC-EPSRC Academic Centre of Excellence in Cyber Security Research at the University of Warwick and a Fellow and Professor of the Alan Turing Institute, where he is principal investigator on a $9 million project developing trustworthy digital infrastructure. Carsten is also the Director for Research and Innovation at EDGE-AI, the UK National Edge Artificial Intelligence Hub and leads security and privacy work on two further large projects funded by EPSRC, leading on security and privacy: the Framework for Responsible AI in Finance project; and Assurance and Insurance of AI.
Anita Khadka is an Assistant Professor at the University of Warwick, with a focus on resilience and assurance. Her research develops foundational and applied methods to address the technical risks and governance challenges of frontier AI technologies, supporting their safe and responsible deployment. She has contributed to major programmes funded by EPSRC, Research England, and Horizon, and has experience securing grants and leading multi-stakeholder projects. Her work has appeared in leading venues, including ACM Computing Surveys, RecSys, and LREC. She is currently leading an EPSRC-affiliated project on the responsible deployment of frontier AI in financial services, integrating technical, legal, and business perspectives to develop sector best practices.