I wonder how long this has been the case?

[Alan Bawden <ALAN@MIT-MC.ARPA>]

In EMACS running on MC read in a file from AI (say AI:COMMON;FILES LOSS).
Then -in that buffer- execute in the minibuffer:
    M.M^R Backward Delete Hacking Tabsâ—Š
The "Undefined or ambiguous function name" error you should get is
apparently caused by emacs going and loading "AI:EMACS;BARE :EJ" instead of
the version of BARE on MC!  (When I translate AI:EMACS; into MC:EMACS; the
problem goes away.)

I don't think this has anything to do with GZ's filename changes.  It looks
like this happens in all versions.