Specifically about generative bots: the bots themselves are no threat. And they would be still no threat, even if they worked far better or far worse than they do - because they simply output text. The actual threat is some bloody irrational (assumptive, gullible, oversimplifying) individuals in positions of power might get really lazy, use the output for decision making, harm people with their idiotic decision, and then blame the tool for what was ultimately their decision. But frankly? This sort of trashy individual has been a plague to humankind since the dawn of time, with or without large language models.