That's something that is explicitly not done, unless there is a very good rationale for why it would need to be prisoners studied. The reason being, it is all but impossible for a prisoner to give a proper informed consent. There is just too large a conflict for testing to be considered voluntary, especially if there is some kind of quid pro quo. See here [0] for a list of permitted research on prisoners (45 CFR 46).

True, it has been done in the past, but there are many... many horror stories associated with the practice [1].

[0] https://www.ecfr.gov/cgi-bin/retrieveECFR?gp=&SID=83cd09e1c0...

[1] https://en.wikipedia.org/wiki/Experimentation_on_prisoners