What say all of you who worked on the AI that powers this?
- Robert J. Oppenheimer
The 'do nothing and don't advance' option is just an illusion. Because in actuality what happens is "Only the worst people get the option.". Life isn't very kind to those attempting moral purity unfortunately. There is a reason 'worldly' and 'morally pragmatic' are synonymous.
If you want to reduce harm, if possible countermeasures are probably the best you can do. It varies by specific technology. Depending upon the tech this may ironically involve working with the very thing you are trying to weaken. Or it may be something completely different, say working in materials science instead of ballistics.