MLCSS'25@UCL

Dr. Andrew Paverd

Principal Research Manager, Microsoft Security Response Center (MSRC).

Bio: Andrew Paverd is a Principal Research Manager in the Microsoft Security Response Center (MSRC), based in Cambridge. He and his team analyze all reported security and privacy vulnerabilities in Microsoft's AI systems and work to develop new mitigations. His research interests also include web and systems security. He received his DPhil from the University of Oxford in 2016.

Talk Title: Vulnerability Response in the Era of AI

Abstract: Coordinated vulnerability disclosure (CVD) is an industry-standard practice to ensure that vulnerabilities can be disclosed, remediated, and published in a coordinated manner. Many companies have well-established vulnerability response programs to support and even incentivize CVD. Drawing on examples from the Microsoft Security Response Center (MSRC), this talk explores how these programs have changed to both use and protect generative AI systems.

Dr. Vasilios Mavroudis

Principal Research Scientist & Co-Lead, AI for Cyber Defence (AICD), The Alan Turing Institute.

Bio: Dr. Vasilios Mavroudis is a Principal Research Scientist co-leading the AI for Cyberdefence (AICD) Research Center at the Alan Turing Institute in London, UK. His work operates at the intersection of systems security and machine learning, with an emphasis on autonomous cyberdefence, AI cyber risk evaluation, and model-based threat simulation. He leads national-scale efforts including the AI Cyber Risk Benchmark and the International AI Safety Report (2024-25), to which he contributed the section on offensive AI capabilities. His work spans academic, policy, and applied security communities.

Dr. Ilia Shumailov

Founder & CEO, Sequrity.ai; Former Senior Research Scientist, Google DeepMind.

Bio: Dr. Ilia Shumailov holds a PhD in Computer Science from the University of Cambridge. Until recently Ilia was a Senior Research Scientist at Google DeepMind focusing on the intersection of machine learning, privacy, and computer security. At present Ilia runs a company sequrity.ai building tools to secure AI agents of the future.

Talk Title: Is AI Security doomed?

Dr. Peter Garraghan

CEO & Founder, Mindgard; Professor of Distributed Systems, Lancaster University.

Bio: Dr. Peter Garraghan is Founder and Chief Science Officer of Mindgard, Professor in Computer Science at Lancaster University, and fellow of the UK Engineering Physical Sciences and Research Council (EPSRC). He is an internationally recognised expert in AI security, Peter has dedicated years of scientific and engineering expertise to create bleeding-edge technology to understand and overcome growing threats against AI. He has raised over £12 million in research funding, been featured in The Register and Forbes, and published over 70 scientific papers.

Talk Title: Test AI Systems, Not Models

Abstract: LLMs leveraged within GenAI applications and agentic systems introduce a big problem in security. They are opaque, non-intuitive, and probabilistic. This makes it particular difficult to understand, model, or assess their security properties. While AI-driven applications unlock new possibilities, their deployment also introduces new vulnerabilities and amplifies existing threats. Much of the attention in AI security has been on security testing models in isolation, but this approach overlooks the bigger picture. Real-world threats do not emerge from models alone, they manifest when AI is integrated into applications and broader systems. This will cover key challenges within AI security, what's new (and the same), show case real vulnerabilities and exploits, and discuss how academic researchers can help push the field forward.