We consider the task of discovering categories of non-pharmaceutical interventions during the evolving COVID-19 pandemic. We explore topic modeling on two corpora with national and international scope. These models discover existing categories when compared with human intervention labels while reduced human effort needed.