Is it inevitable that LLMs will soon develop their own language, leaving humans in the dark?

It seems relatively uncontroversial to conclude that human language, which has evolved exclusively to deal with human issues, is not the optimal means of communication between AI systems. Given AI's ever increasing ability to optimize their own processes through self-learning, AI systems developing a more suitable language seems within the realm of possibility. If it proves to be more efficient as well and reduce cost, the laws of free market competition dictate that it will happen unless explicitly prevented. What do you think this would entail, and should we attempt to take measures to prevent it?