While it's not exactly the same as taking control of a nuclear arsenal and firing off missiles to wipe out humanity so as to bring forth robotic supremacy, this is still pretty scary.
A research program at Facebook created two artificial intelligence systems with the goal of having the two communicate negotiate with each other. Eventually, the two AIs began communicating with one another in a broken form of English, whereby they dropped words that weren't necessary for communicating with each other. This meant that it was more and more difficult for the researchers observing them to understand.
While there's nothing dangerous - so far, anyway - of AIs creating their own language, it does present a problem in that their language could be seen as more efficient than what humans have used for centuries. More to the point, if AIs learn to speak to each other in codewords that blocks out human understanding, then it's getting somewhat troubling.
The ultimate goal of the research program was to build an AI that could talk to and understand humans, but the fact that the AI created its own language - all by itself - is something that nobody really expected.
Either way, we're still a long way away from Skynet or HAL 9000.