This is just a guess, but maybe those are papers that are cited in others from MS Research authors, and there was a data migration bug at some point that mixed citations with authored papers.
It is obviously a bug since MSR was founded in...1991. Many researchers also insert papers into the system written in other labs or universities before joining MSR; these seem to be included as well even though these days we clearly get to distinguish between was written when at MSR and what wasn't. This behavior seems to only apply to older papers and not newer ones.
Nice find. Sorting by oldest first turns up pages and pages of classic papers, e.g. by Kolmogorov, Bucy, David Marr, etc., etc. Many of these people have never had a relationship with MS. And they are, in many cases, copyright by other entities, like the IEEE or the Royal Statistical Society, so they are in no way MS IP.