We produce work that bridges technical AI safety and governance — the kind of analysis that is legible to both an engineer and a policymaker.
We do not produce position papers for their own sake. Our focus areas — AI safety, AI governance, and the philosophy of AI — are not treated as separate disciplines. The most important questions sit at their intersections, and that is where we work.
Our research is designed to speak across audiences. A paper on alignment risk should be as useful to a legislator as it is to a machine learning researcher. A governance proposal should be grounded in a real understanding of how the systems actually work. This is harder than working within a single discipline, but it is where the most important contributions will come from.
Technical and conceptual research on ensuring AI systems behave as intended and do not cause unintended harm. We are interested in alignment, robustness, interpretability, and the challenge of specifying what we actually want from increasingly capable systems.
Research on the regulatory, institutional, and policy frameworks needed to govern AI systems responsibly. We focus on how governance can keep pace with fast-moving technology, and how it can be informed by genuine technical understanding rather than abstraction.
Foundational questions about the nature of intelligence, agency, consciousness, and moral status as they apply to artificial systems. We think the philosophical questions are not downstream of the technical ones — they shape how we understand what is being built.
Our first research outputs are in development. Subscribe to our mailing list to be notified when we publish.
Ensemble