Why do you care about AI Existential Safety?

I care about AI existential safety because I’ve always been sensitive to questions of responsibility, control, and moral accountability. As systems become more powerful and more opaque, I worry about a future where meaningful decisions are made without clear ownership or the ability to intervene, and where humans slowly step back from responsibility. That possibility makes me uneasy, because it suggests a future where loss of control is inevitable and we might begin to face existential risks from AI sooner rather than later.

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

My research focuses on scalable oversight and deployment-time safety mechanisms. I study how to combine powerful models with more conservative safeguards, using formal risk control to limit unsafe behavior while preserving usefulness. This approach is motivated by the concern that future AI systems may be too complex or capable for direct, continuous human supervision.

By designing mechanisms that decide when to trust a model, when to intervene, and how much autonomy to grant over time, my work aims to prevent small alignment failures from compounding as systems scale. If advanced AI systems act in ways we cannot reliably monitor, correct, or constrain, loss of human control becomes increasingly likely, greatly increasing existential risks. ​