This seems to be just a feature of katago, not a feature of all Go AIs. If this also works against implementations of AlphaGo, then you can start to claim that there are no superhuman go AIs.
This seems to be just a feature of katago, not a feature of all Go AIs. If this also works against implementations of AlphaGo, then you can start to claim that there are no superhuman go AIs.