Several things were wrong here:
Firstly, the gems had been installed as root, which meant that their codebase (and any shell scripts or binary commands their installation generated) were root's, not the current users. This was the cause of OP's original failure - namely, the executable scripts could not resolve the gems since they were installed under a different user's rvm environment.
Secondly, even had the gems commands been accessible this could have created all sorts of security problems, given that the files belonged to root. Potentially, if these gems had contained badly-behaved code (or malicious code) embedded within them, a privilege escalation attack might have been possible against the system running the ruby process.
Basically, the TL;DR is that you should always install the gems using the credentials of the user who needs to run the ruby process that requires the gems, and this user should never be root
or have superuser rights to the system or filesystem.