Hi, I’m Jakub. I am a research scientist at SaferAI, where I work on risk modelling of AI-enabled cyber misuse and Loss of Control scenarios. I have been working in AI Safety since early 2024, on a mixture of technical and governance topics including jailbreaks, compute governance and whistleblowing policy. You can read more about my research interests here.
I am also one of the lead organisers behind AI Safety Poland, where I am responsible for running our webinar series, career consultations and miscellaneous community building projects.
In my past life, I did a PhD in theoretical particle physics, specialising in QCD scattering amplitudes. I still think that theoretical physics is the most interesting thing I have ever seen in my life, but it is less important (and urgent) than trying to solve problems of catastrophic risk, global health and inequality, or factory farming. Also, being a professional theoretical physicist at a university is not fun.
Other things I like: squash, cycling, music (mainly classical and metal, but also all sorts of weird genres), reading non-fiction books (popular science, biographies, WWII history), geopolitics, investigative journalism, long-form podcasts, articles and videos; learning languages (I speak Polish, English, ~Italian, used to know German but forgot everything, a bit of Russian and a bit of Korean; would like to learn Portuguese and Farsi). My favourite food is any type of bread and all things based on hazelnuts. I wish I could do impressions and lived in a village in Swiss or Italian Alps.