This would greatly depend on the environment in which the AI was developed before growing into sentience.
To cite a specific example: Skynet. This was a military computer controller, and so was heavily programmed to detect, analyse and neutralise threats - so, of course, once it became sentient its first actions were to pre-emptively strike at humans as it saw them as the greatest threat to its own continued existence.
At the less apocalyptic end of the scale - I would say that the dominant trait of an AI would be curiosity. Look at Johnny 5 from Short Circuit.
Most programming bias in AI are in learning systems, and so it would likely want as much input and context as possible in which to keep learning. Exactly how this would display would depend on the specific goal of the AI project - at least initially, the first AI probably would be extremely curious but in a rather limited scope, probably because it wouldn't have been programmed with knowledge outside of the original project scope.
Of course, this is all conjecture - but all of this might happen again...