Abstract

Abstract: As LLMs become part of products used daily by millions of people, it is increasingly urgent to ensure that these models are developed and operate responsibly. In this talk, I will share my journey in Responsible AI (RAI), from my early work in automated fact verification to the development of the Amazon Nova family of foundational models and my new role in the Information Commissioner’s Office (ICO), the UK’s data protection regulator. I will share some examples of RAI research that I and my former team have done, including the largest RAI benchmark for video foundation models, and a linguistically-inspired text-to-text LLM watermarking method. I will also talk about the role regulatory agencies like the ICO play in the extremely fast-paced world of post-ChatGPT AI. Finally, I am going to discuss how RAI is practiced in an industry setting and how it is influenced by and sometimes informs AI regulations.

Bio: Dr Christos Christodoulopoulos is a Principal Technology Adviser for the Information Commissioner’s Office, the UK’s data protection regulator, working with the AI Policy and AI Compliance teams. Before that, he was a Senior Applied Scientist at Amazon, working on Responsible AI for Alexa and the Amazon Nova family of Foundational Models. He has a PhD in Computational Linguistics from the University of Edinburgh, and he was a postdoctoral researcher at the University of Illinois. He is a Program Chair for EMNLP 2025 , a Senior Area Chair for various *CL conferences and was the General Chair for the 2021 Truth and Trust Online conference.