Dr Christian Schroeder de Witt - University of Oxford

Research Fellowships 2025

The rapid advance of generative AI is transforming every layer of society, from how we access information to the way autonomous systems, individuals, and institutions such as governments interact. As these AI agents and networks increasingly collaborate and even outmanoeuvre traditional oversight, new and complex security challenges arise – threats that cut across technical, organisational, and societal boundaries. Synthetic content now proliferates at unprecedented scale, fuelling misinformation, enabling covert communication, and threatening the integrity of critical infrastructure – from autonomous driving to the emerging “AI internet”. Meeting these urgent risks demands a genuinely interdisciplinary and holistic response. Dr Schroeder de Witt’s Fellowship will unite expertise from mathematics, computer science, psychology, cybersecurity, and policy, partnering with world-leading organisations such as Adobe, Armasuisse, the BBC, and the UK Intelligence Community. Together, we bridge the technical and human factors, developing real-world solutions.

Dr Schroeder de Witt’s unique contribution is a mathematical-grounded, anticipatory security perspective: developing rigorous models to predict, characterise, and mitigate AI-driven threats before they escalate. By pioneering security-by-design methods grounded in information theory and statistical learning, Dr Schroeder de Witt will create tools to detect deepfakes, out-of-context misinformation, and steganography (concealed content). These solutions will be explainable, scalable, and robust, transcending the limits of existing oversight mechanisms. These advances are designed to complement and integrate with content provenance systems, psychological research, and human-centered safeguards, ensuring that technology, standards, and societal values reinforce each other. 
 
Rather than being reactive, this Fellowship aims to anticipate and pre-empt threats at the system level, founding a new discipline of multi-agent security for the generative AI era. By integrating diverse expertise and a holistic viewpoint, Dr Schroeder de Witt aim to strengthen digital trust, safeguard public discourse, prevent financial fraud, and enable the safe, responsible adoption of AI throughout society.

Personal website

LinkedIn

X (Twitter)