We will discuss this recent paper which is related to the “Gradual Disempowerment” paper from about a year ago: https://gradual-disempowerment.ai/ There is also an interesting podcast interview with one of the authors.
Daniel Tan is one of the main authors of the Emergent Misalignment paper and also the main author of a paper introducing Inoculation Prompting. In many of our discussions so far, we have talked about these papers, so this could be a very interesting event for many of you!
We will discuss the paper Legal Alignment for Safe and Ethical AI by Kolt et al. (2026).
We will discuss the paper The Artificial Self: Characterising the landscape of AI identity by Douglas et al. (2026).
Yannick Mühlhäuser is working as a Policy Researcher at the Future of Life Institute and as Research Manager at ERA Cambridge. Previously, he was a Talos Fellow in Brussels and studied Physics in Tübingen. Yannick will briefly give an overview of his work on hardware-enabled governance and AI verification, followed by an open Q&A. Since […]
Johannes will give a short overview of results from the Mythos Preview System Card, and then the main focus of the meeting will be on discussion: Does this change how models will be released from now on? What are the safety and security implications? Should this update us on timelines or specific threat models? If […]
Test your knowledge of AI facts while practicing calibrated uncertainty, in a fun estimation game by Markus Anderljung (https://www.markusanderljung.com/ai-calibration-game.html). We will split into small groups, and compete for the highest score. Everybody is welcome to join! Since many of you will be busy with NeurIPS submissions this week, we are aiming to keep this meetup […]
Alexander Panfilov (Sasha) will present his recent work on automating the discovery of adversarial attack algorithms for LLMs (https://arxiv.org/abs/2603.24511). This paper has gotten quite a bit of attention, as the results are an early demonstration that incremental safety and security research can be automated using LLM agents. We are happy to host Sasha for a […]
We are hosting a Q&A about the MATS fellowship with current and previous fellows. We’ll cover what it’s like to do research at MATS, give some advice for the application process, and try to answer all your questions about the program. You can find more information here: https://www.matsprogram.org/program/autumn-2026 As usual, we’ll provide dinner and drinks.
