Alex Csaky

Email · LinkedIn · Substack

I'm an AI safety researcher and founder working on AI for epistemics and coordination: how AI can help people form better beliefs, make wiser decisions and coordinate. I'm currently an Astra Fellow at Constellation in Berkeley, focusing on AI and decision-making: how AI can help people and institutions make better decisions, and how AI systems make their own.

I also mentor a SPAR project on AI for epistemics, have just co-authored State of the Field: AI for Epistemics and Coordination, and consult on enterprise AI transformation through Proferra.

Previously: contracted with Forethought as a founder in residence, scoping the field of AI for epistemics and coordination; Talos Policy Leaders Fellow (EU AI governance); Apart Studio Fellow (sycophancy evals); facilitator on BlueDot's AGI Strategy course. Before AI safety I founded and sold an e-commerce startup while at secondary school, ran an education company in Hong Kong, and worked as an AI engineer building tools for enterprises and political candidates. I read archaeology and anthropology at Oxford, where I dug on a Roman Iberian excavation at Castelo de Cuncos and wrote my dissertation on human cognitive enhancement.

Things I'm interested in

• AI for epistemics and coordination. State of the Field: AI for Epistemics and Coordination (report on LessWrong, October 2026), building on my scoping work with Forethought.
  • AI and decision-making. Through my Astra Fellowship: how AI can help people and institutions make better decisions, and how AI systems, including groups of AI agents, make their own.
  • Mentoring. AI for epistemics: decision-making evals, deployment research and cause prioritisation, a SPAR project (Fall 2026).
  • Multilingual model evaluation. Syco-Lingual (short paper; presented at IASEAI'26 Paris).
• Hosting. I co-run the AI Safety Cinema Society in London; our first screening paired Daniel Roher's The AI Doc: Or How I Became an Apocaloptimist with Suzy Shepherd's short Seat at the Table. I organise workshops and hackathons. Beyond work, I like hosting dinners and events for friends.
• Anthropology and Futurism. I'm writing an essay on AI personhood for Scheme magazine, arguing that personhood is made through relationships rather than crossed as a threshold (semi-finished draft). More widely I enjoy reading, writing and thinking about the future of humanity, as well as anthropology and what it means to be human. See the Reading / Writing page for more of this.