I don't understand if the AI knew that by passing a move it would lose or if it thought that all of its supposed territory actually belonged to itself. In the first case the adversary really won the game. The second case is something that can happen between humans, especially beginners. If the players do not agree on the result they keep playing until they do, with some adjustments depending on the flavor of the rules (basically the usual area vs territory scoring.)
KataGo was trained on multiple scoring methods at the same time: it is an input to the algorithm[0]. The model learnt that it would win when passing, and it seems it never had the opportunity to detect that it would not win under Tromp-Taylor when passing, because its opponent in self-play, KataGo, then either passed and lost (under other rules) or resigned.
So it's a bug in the training method, probably a very minor one because nobody already exploited it. The only really interesting thing in here is that it took another AI to find that bug.
