Phonetic analysis of the reflexes shows that both EC and WC languages reflect two initial variants (with voiceless *-t- and with glottalized *-ṭ-), and it is hard to decide which is the original one. All other correspondences are surprisingly regular (if we separate all the secondary interlingual borrowings). The etymon is, of course a Wanderwort (cf. Kartv. *ḳaṭu, late Latin cattus, Slav. *kotъ, Osset. gädɨ, see Abayev 1958, 510, etc.). Klimov (1967, 380-381) tried to explain its presence in the Caucasian languages as a loan from Armenian; but the word no doubt existed already at least in PWC and in sub-protolanguages like PL and PA which makes this theory impossible.
Even if the word is borrowed from some external source, the borrowing occurred so early that we think it justified to include the word into the common NC wordstock.