Why do you care about AI Existential Safety?

I care about AI existential safety because I want to help protect humanity’s long-term future. My experiences, from seeing how fragile social systems can fail vulnerable people to later living in a highly cooperative international community at United World College, made me care deeply about what allows complex systems to remain trustworthy, stable, and aligned with human flourishing. That concern led me to effective altruism and to AI safety, where the stakes seem especially high. I believe advanced AI could pose existential risk if developed without adequate safeguards, and my research has made me especially concerned by how subtle vulnerabilities in multi-agent systems can scale into systemic harm. I am also worried that current safety evaluations often lack the rigor required for such high-stakes systems, because if we cannot reliably measure safety, we risk deploying systems that only appear safe.

Please give at least one example of your research interests related to AI existential safety:

My research interests focus on developing principled methods, reporting norms, and evidence standards for studying emergent safety-relevant behaviors in advanced AI systems. I am especially interested in multi-agent settings, where behaviors such as deception, collusion, strategic adaptation, and coordination failure are more likely to emerge and harder to measure reliably. Because these dynamics are rare, context-sensitive, and easy to misinterpret, I aim to help build evaluation methods that produce robust, decision-relevant evidence rather than fragile results or false confidence. My broader goal is to connect this technical work to AI governance, so that safety research can better inform which risks are technically manageable, which require institutional regulation, and how society should respond to transformative AI.