I study how to make AI systems secure and reliable, and how to evaluate whether they behave that way in practice.https://mitchellhermon.com/en-us