Securing the Age of AI Agents: An Industry Briefing

An Industry Briefing on Frontier AI Safety, Security, and Privacy Research

Join us for an in-depth look at frontier research in AI Safety, Security, and Privacy.

As AI systems are moving beyond task-specific models toward autonomous agents capable of taking actions, writing code, and interacting with other systems, the security, privacy, and safety challenges are changing just as rapidly. Capabilities that once sounded like science fiction are quickly becoming engineering reality: from AI writing its own exploits and autonomously patching vulnerabilities to collaborating with other AI agents.

This industry briefing brings together leading researchers and practitioners at the forefront of AI cybersecurity to explore what is changing, what risks and opportunities lie ahead, and what organizations need to prepare for.

Hosted at KU Leuven, this event offers a unique opportunity to hear from leading academic researchers with active ties to industry: from Google Research and Microsoft to AI cybersecurity startups and XBOW.

The afternoon will explore three interlocking themes:

• Safety alignment and benchmarking — How can we evaluate whether modern AI systems behave as intended, and where do current benchmarks fall short?
• Security and privacy of AI agents — The new attack surface created by agentic AI, from data leakage to manipulation of autonomous decision-making, viewed from both defensive and offensive perspectives.
• Beyond the hype — Separating evidence from narrative around agentic AI's role in autonomous vulnerability discovery and cybersecurity economics.

PROGRAM & SPEAKERS

13.00: Registration & Welcome coffee

13.30: Opening remarks

13.40: Sahar Abdelnabi: AI Agents May Always Fall for Prompt Injections****.****
(Principal Investigator at ELLIS Institute Tübingen & Max Planck Institute for Intelligent Systems; formerly Microsoft)

14.25: Eugene Bagdasarian: Security & Privacy of AI Agents.
(Assistant Professor, UMass Amherst; Senior Research Scientist at Google)

15.10: Break

15.40: Brendan Dolan-Gavitt: AI Agents for Offensive Security
(AI Researcher at XBOW; former Associate Professor at NYU Tandon School of Engineering)

16.25: Lorenzo Cavallaro: Demystifying the Mythos or Disrupting Bugonomics?
(Full Professor of Computer Science at University College London; Chief Scientific Officer at Bynario, an AI-driven vulnerability discovery startup)

17.25: Discussion Panel: Moderated Q&A with a panel of speakers and invited experts

18.30-20.30: Networking Reception

PRACTICALITIES****:****

Venue address: Celestijnenlaan 200A, 3001 Heverlee | Computer Science dept. | Aula Duval

Parking Kapeldreef - access code: 11063#

** This industry event is supported by the Cybersecurity Research Program Flanders.