The bug move (a rook) was in game 1. It was inconsequential to the game result. The position was already lost for Deep Blue. It made a random move because everything had the same outcome of losing. The bugged move simply made it lose sooner. Kasparov knew this and there was no controversy.
In game 2, Kasparov set a trap, offering a sacrifice of two pawns for great positional improvement. Kasparov thought Deep Blue couldn't calculate far enough ahead to see the positional improvement, but it did. This made Kasparov suspect and accuse IBM of human intervention in the computer's moves.
The fatal move (a knight sacrifice by Deep Blue) was in game 6. Kasparov deliberately played an opening with a known weakness, that he thought the computer wouldn't find as it wouldn't sacrifice material and calculate far enough ahead to see the positional advantage. Deep Blue found it. Kasparov again accused IBM of human intervention. The Deep Blue team then said they had added this variation into its opening database just before that game, pre-analyzing that line deeper since Kasparov had previously played it in game 4. There was some dispute as to whether the rules for the competition were intended to allow IBM to modify Deep Blue during the match.
The bot hit a bug that chose a strange move, and since kasparov couldn't figure out what the line was, he resigned. If he analysed the board instead of what the bot was thinking, he would have won easily