PhD in AI Security and Information-Theoretic Foundations of Memory Poisoning Linköping University in Sweden

Linköping University is advertising a PhD position in AI Security focused on the information-theoretic foundations of memory poisoning. The project studies how malicious information injected into an AI agent’s memory can affect reasoning and decision-making, with an emphasis on rigorous theory as well as proof-of-concept attacks, defenses, and experiments with LLM-based agents.

The position is based in the Division for Communication Systems at the Department of Electrical Engineering, Linköping University. The research team spans Linköping University, Chalmers University of Technology, and Recorded Future, and includes work in AI privacy and security, communication theory, statistical learning, and decentralized machine learning.

Research keywords: AI security, information theory, memory poisoning, LLM agents, privacy, statistical learning, communication theory, signal processing, optimization, applied probability and statistics, cybersecurity, distributed systems.

Eligibility: Applicants should have a Master’s degree in Computer Science, Electrical Engineering, or Applied Mathematics (or equivalent), with at least 240 credits and at least 60 advanced credits in the relevant subjects. Strong mathematical maturity, programming skills, and excellent English communication skills are required. Prior experience in AI security, privacy, or LLMs is helpful but not mandatory.

Funding: The position is funded by the Swedish Research Council (VR). The employment is normally four years full-time equivalent, with possible extension up to five years depending on teaching and departmental duties. The starting salary is around 36,400 SEK/month, with employee benefits including a pension.

Application deadline: 2026-06-22. Apply through the Linköping University vacancy page.