protein folding simulations don't produce a lot of money for pharma (I say that as somebody with a F@H publication to my name specifically doing research on GPCRs, which are a huge pharma target). users can see what the workunits are doing, and choose subprojects, presumably avoiding bioweapons if that's important to them.
How could the average technically savvy user, let's say willing to use Google a bit but not to obtain a biochemistry degree, tell which subprojects are possible bioweapons? Not a rhetorical question.
I'm going to guess that if you inspect each project it has a purported (positive) medical use and you'd have to make a judgement call on each project based on good knowledge of biology as to whether it had a bioweapons application.
I doubt, but cannot say for certain that F@H won't have any projects that are stated as bioweapons and I'm skeptical that somebody working on a bioweapon would try to cloak its intent, and come up with a plausible medical application so they could get free F@H cycles.
I remember the days of logging into national labs supercomputers and seeing the other jobs on the system, such as job_name "submarine_geometry_optimizer" and the user_name was "electric_boat" or something similar.
