Representation Engineering · AI Monitoring

Maciej Chrabąszcz

How to say my name

Maciek MA‑check

  1. MA “ma” as in mama.
  2. checkexactly like the English word check; the “ch” is soft, as in cheese.

Maciej (MA‑chey) is the formal form you will see on my papers. Maciek is the everyday one, and the one I prefer.

Chrabąszcz hra‑BONSH‑ch

  1. hra a breathy “h”, as in loch, straight into “ra” as in the sun god Ra.
  2. BONSH “bon” as in James Bond, then “sh” as in shoe.
  3. ch as in church.

The “szcz” ending is just “sh” + “ch”, like the middle of fresh cheese.

Fun fact: Chrabąszcz means “May beetle” in Polish. 🪲

I study the hidden representations of generative models: how to read them, steer them and monitor them, and how that can help make these systems safer.

PhD candidate, Warsaw University of Technology
AI Safety Researcher, NASK National Research Institute

Maciej Chrabąszcz

About

I am a PhD candidate at Warsaw University of Technology and an AI Safety researcher at NASK National Research Institute. My thesis, supervised by Prof. Tomasz Trzciński and Dr. Sebastian Cygert, is on efficient machine-learning methods for the safety of AI models.

At NASK I co-led the safety and evaluation work for PLLuM, the Polish government's family of Polish large language models. I hold an MSc in Mathematical Statistics and Data Analysis from WUT.

Research interests

I'm interested in AI safety in general. My work currently centres on white-box methods for AI monitoring and steering: using a model's internal activations to build cheap guardrails, track its reasoning, or change its behaviour. I also work on safety evaluation of LLMs.

Publications

Google Scholar

* denotes equal contribution.

2026

2025

2024

2023

Get in touch

Happy to talk about AI safety, representation engineering or possible collaborations. Email is the best way to reach me.