EDIC

The seminar will explore recent research on the security of AI agents
and agentic systems. In particular, we will look at approaches based
on programming languages, compilers and runtimes, capability-based
security, and related techniques for controlling and reasoning about
the authority of AI agents. We will also cover relevant background,
such as the object-capability model, and discuss concrete agent
systems and security mechanisms.  

The seminar will be based around paper presentations and discussion,
with students presenting selected papers.

Please check back for updates. The reading list and schedule will be made available here shortly.