Egor Zverev

Hi! My name is Egor and I am a PhD student at ISTA, Austria, working under the supervision of Christoph Lampert. I am also a member of the ELLIS PhD Program co-supervised by Florian Tramèr from ETH Zürich. Since June 2026, I have been a research intern at Microsoft Research Cambridge, working with the Project Roma team on secure-by-design web agents, and I also advise several master's students on AI security in robotics through the ETH Robotics Club.

I am broadly interested in AI Safety and Security, with a particular focus on improving LLM Security through architectures and building provable safeguards for AI agents. In my previous work I have formalized the problem of instruction-data separation (i.e., what it means for the model to separate executable instructions from non-executable data) and proposed a method to increase such separation through architectural changes. More recently, I have been working on provable security guarantees for web agents that need to handle untrusted content.

Before coming to ISTA, I got my B.S.(Hons) in Applied Math and CS from the Yandex Department of Data Analysis at the Moscow Institute of Physics and Technology, where I also taught Python and stats. You can find my full CV here.

I also enjoy doing creative and community-driven gigs. I was the lead organizer for the Foundations of LLM Security Workshop @EurIPS'25 and the LLM Safety and Security ELLIS UnConference'25, and I am now co-organizing the 2nd iteration of the Foundations of LLM Security Workshop @NeurIPS'26. I co-created one of the AI art booths for ISTA Summer Campus, and in my spare time, I write poetry and tinker with Arduino.

I am always open to new connections, professional and otherwise, feel free to send me an email at [first_name].[last_name]@ist.ac.at. Let's chat!