Most of them are full time researchers publishing papers. For example someone might do a study on facial recognition accuracy on different ethnic groups and how that impacts user experience or other disparate impact of automated systems.
Apparently they do nothing like you'd expect from something called "Ethical AI group".